prevención en el acoso

31
Prevention & Treatment, Volume 6, Article 21, posted October 1, 2003 Copyright 2003 by the American Psychological Association Predicting Teachers’ and Schools’ Implementation of the Olweus Bullying Prevention Program: A Multilevel Study Jan Helge Kallestad University of Bergen and Bergen University College Dan Olweus University of Bergen ABSTRACT Little is known about factors that predict or affect differences in teachers’ and schools’ implementation of school-based intervention or prevention programs. The main purpose of the present project was to study this important issue in a sample of 37 schools and 89 teachers who provided data at 2 points in time, separated by 6 months. All of the teachers used, to varying degrees, the Olweus Bullying Prevention Program, which has been shown to reduce substantially bully/victim problems in school. Two measures of implementation were constructed, the Classroom Intervention Measures (CIM) and Individual Contact (with bullied/bullying students and/or their parents; IC). Generally, substantial amounts of variance in implementation could be predicted in multilevel models. In the 2 within-school models, 5 teacher-level factors predicted 53% (CIM) and 34% (IC) of the variance, respectively. Two of the predictors, Perceived Staff Importance and Perceived Level of Bullying (in own class), were common to both models. At the school level, only CIM showed systematic between-school variance; in the between-school model, 3 school-level predictors accounted for 50% of the variance. The school climate measures Openness in (Staff) Communication and School Attention to Bullying Problems were both important predictors. In summary, the meaningful results from our study indicate that teachers were the key agents of change with regard to adoption and implementation of the Olweus Bullying Prevention Program in school. Generally, we think our study has shed light on several factors of importance and has contributed to a better understanding of the process of program implementation. The empirical results have also This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

description

prevención en acoso

Transcript of prevención en el acoso

  • Prevention & Treatment, Volume 6, Article 21, posted October 1, 2003 Copyright 2003 by the American Psychological Association

    Predicting Teachers and Schools Implementation of the Olweus Bullying Prevention Program: A Multilevel Study

    Jan Helge KallestadUniversity of Bergen and Bergen University College

    Dan OlweusUniversity of Bergen

    ABSTRACTLittle is known about factors that predict or affect differences in teachers and schools implementation of school-based intervention or prevention programs. The main purpose of the present project was to study this important issue in a sample of 37 schools and 89 teachers who provided data at 2 points in time, separated by 6 months. All of the teachers used, to varying degrees, the Olweus Bullying Prevention Program, which has been shown to reduce substantially bully/victim problems in school. Two measures of implementation were constructed, the Classroom Intervention Measures (CIM) and Individual Contact (with bullied/bullying students and/or their parents; IC). Generally, substantial amounts of variance in implementation could be predicted in multilevel models. In the 2 within-school models, 5 teacher-level factors predicted 53% (CIM) and 34% (IC) of the variance, respectively. Two of the predictors, Perceived Staff Importance and Perceived Level of Bullying (in own class), were common to both models. At the school level, only CIM showed systematic between-school variance; in the between-school model, 3 school-level predictors accounted for 50% of the variance. The school climate measures Openness in (Staff) Communication and School Attention to Bullying Problems were both important predictors. In summary, the meaningful results from our study indicate that teachers were the key agents of change with regard to adoption and implementation of the Olweus Bullying Prevention Program in school. Generally, we think our study has shed light on several factors of importance and has contributed to a better understanding of the process of program implementation. The empirical results have also

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • suggested ways in which implementation of the program can be improved, and several of these amendments have already been incorporated in the program and its dissemination (Olweus, in press).

    This research was conducted as part of Jan Helge Kallestads fulfillment of the requirements for a doctoral degree of philosophy at the University of Bergen, Norway, under the supervision of Dan Olweus. The research reported in this article was supported by grants from the Norwegian Research Council and from Rdet for psykisk helse to Dan Olweus Jan Helge Kallestad, grants to Dan Olweus from the Ministry of Children and Family Affairs and the Norwegian Public Health Association (Nasjonalforeningen for folkehelsen), and, in earlier phases, grants to Dan Olweus from the Ministry of Education, which is gratefully acknowledged. Correspondence concerning this article should be addressed either to Jan Helge Kallestad or Dan Olweus, Research Center for Health Promotion, University of Bergen, Christies Gate 13, N-5015 Bergen, Norway. E-mail:[email protected] or [email protected]

    A strong societal interest in bullying or bully/victim problems among chool children was first aroused in Sweden in the late 1960s and early 1970s. This interest resulted in a fair amount of public attention in Scandinavia and initiation of what is generally regarded as the first systematic research project on the phenomenon. The project was published as a book first in Sweden (Olweus, 1973) and a few years later in English under the title Aggression in the Schools: Bullies and Whipping Boys (Olweus, 1978).

    In Norway, bully/victim problems were for several years an issue of general concern in the mass media and among teachers and parents, but the school authorities did not involve themselves officially with the phenomenon. In the early 1980s, a marked change took place. More precisely, in late 1982, a newspaper reported that three 1014-year-old boys from the northern part of Norway had committed suicide, in all probability as a consequence of severe bullying by peers. This event aroused considerable uneasiness and tension in the mass media and the general public. It triggered a chain of reactions, the end result of which was a nationwide campaign against bully/victim problems in Norwegian comprehensive schools (Grades 19, corresponding to Grades 210 in the current grade system) launched by the Ministry of Education in the fall of 1983. In that context, what has later become known as the Olweus Core Program against Bullying and Antisocial Behavior (Olweus, 1993a, 2001b), or the Bullying Prevention Program (Olweus & Limber, 1999), was developed and systematically evaluated.

    The effects of this intervention program were evaluated in a research project involving 2,500 boys and girls in 42 primary and lower secondary/junior high schools in the town community of Bergen, during the 2-year period from May 1983 through May 1985. At the time of the first measurement, the students were enrolled in 112 classes in Grades 58 (according to the current grade system designation), with modal ages of 11, 12, 13, and 14 years. The positive results of the intervention have been reported in several publications (e.g., Olweus, 1991, 1993a, 1993b; Olweus & Alsaker, 1991; Olweus & Limber, 1999). These results included the following: (a) substantial reductionsby 50% or more in most comparisonsin students reports of bullying and victimization; (b) marked reductions in general antisocial behavior such as vandalism, shoplifting, truancy, and alcohol use; and (c) significant improvements of the social climate of the classroom, as reflected in students reports of improved order and discipline, as well as a more positive attitude toward schoolwork and the school.

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • The results of the first systematic intervention project against bullying were thus very encouraging, with key findings having later been largely replicated and expanded in two new studies in Norway (Olweus, 1999a, 2001a; Olweus & Limber, 1999) and in similar partial replication projects in England (Smith & Sharp, 1994), the United States, and Germany (see Olweus & Limber, 1999, for an overview).

    The Focus of the Present StudyIt is important to realize that the effect findings described earlier represent a kind of aggregate result, reflecting the overall effects of an intervention package with several different components (see the next paragraph and Table 1). To get more detailed information about the nature and mechanisms of these effects, there are several issues that need to be pursued and researched in more detail.

    Table 1Overview of the Olweus Bullying Prevention Program

    General prerequisites Awareness and involvement on the part of adults Measures at the school level

    Questionnaire surveySchool conference dayEffective supervision during break times Staff discussion groupsFormation of coordinating group

    Measures at the class levelClass rules against bullyingClass meetings with studentsMeetings with parents of the class

    Measures at the individual levelSerious talks with bullies and victimsSerious talks with parents of involved studentsTeacher and parent use of imagination Development of individual intervention plans

    A major issue concerns factors that are of importance for the adoption and implementation of the program. A natural precondition for any effects to occur is of course that the program or essential parts of it have actually been implemented and maintained. This issue can be divided into two separate but related subissues. One concerns the general (main effect) conditions or factors that affect adoption and implementation of a program or educational innovation. There exists a good deal of research on this subissue, in particular in the educational field and with a special focus on large-scale educational reforms (e.g., Fullan, 1991). Because of the aims and design of our project (described next), we do not have much research data to shed light on this subissue. Nonetheless, some of the generalizations made by researchers in the educational field will

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • prove meaningful and relevant in the present context and are briefly commented on in the discussion of our results.

    The other subissue, and the one we mainly address in this article, is concerned with differences or variability in implementation of the program. As expected, some teachers and schools clearly implemented more of the program than others. What were then the characteristics of the teachers/schools that can predict or explain these differences? The main concern of the present study was with the teachers/schools responses to a specific, circumscribed intervention program offered to the schools/teachers at a particular point in time. There seems to exist much less research on this subissue to relate to and to build upon. To the best of our knowledge, empirical, quantitative studies of factors predicting differences in implementation of a circumscribed intervention program in the personal/social-development area are very scarce.

    It is obvious, however, that systematic research on this subissue can presumably be very useful and, in particular, can help make implementation of a program or educational innovation more effective. Such knowledge may also contribute to the establishment and development of a science of effective implementation. As emphasized by Biglan (1995), The adoption of an effective practice is itself a behavior in need of scientific research (p. 15).

    Given the absence of a developed implementation science, our empirical analyses in the present article had to be guided and structured by some general expectations about factors affecting implementation derived from prior theorizing and empirical findings. These general theoretical expectations and their bases are briefly outlined in the next section.

    The present article sketches first our general theoretical expectations, which is then followed, in the Method section, by a summary description of the intervention program and its goals. After having presented the key components of the program at the school, classroom, and individual levels, we introduce the two measures of degree of implementationthe Classroom Intervention Measures and Individual Contact indicesalong with a brief description of the selected predictors. We use multilevel techniques for most of our statistical analyses because of the hierarchical structure of the data and because of our interest in identifying factors related to implementation at both the teacher/classroom and the school levels. The empirical analyses are followed by a discussion of the findings and some of their implications.

    Specifically, the key research questions of our study can be summarized in the following way: What factors, if any, can predict degree of or differences in implementation of the Olweus Bullying Prevention Program at the teacher/classroom level? Additionally, what factors, if any, can predict differences in teacher implementation aggregated at the school level (school differences), over and above what has been predicted at the teacher/classroom level?

    General Theoretical ExpectationsPerceived need for a program. As emphasized by Fullan (1982, 1991, 1998), among others, successful implementation of a new intervention program or educational reform is probably possible only if it meets a clear need in the targeted persons or institutions. In the present study, we assumed that a teachers experienced need for a program was related to his or her Perceived Level of Bullying in his or her own classroom, which was one of the predictors of degree of implementation (see the Method section).

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Teacher efficacy. As Smylie (1990) has shown, the concept of teacher efficacy (cf. Bandura, 1986) has been defined in several different ways, but a common element is the beliefs that individual teachers hold about their own capacities or abilities to act in ways that bring about student learning and development (p. 49). Despite the conceptual ambiguities, Research has identified a number of significant relationships between teacher efficacy and teacher work performance and outcomes (Smylie, 1990, p. 57), including implementation of educational innovations (Ghaith & Yaghi, 1997; Hoy & Woolfolk, 1993; McLaughlin & Marsh, 1990). In light of this research, we used a measure of Perceived Staff Importance (with regard to bullying) as a possible indicator of teacher efficacy and a predictor of program implementation.

    Appropriate skills. In general, it is reasonable to assume that implementation of a program is facilitated if teachers have the appropriate skills for dealing with the targeted problem (cf. Calhoun & Joyce, 1998; Fullan, 1991; Green & Kreuter, 1999; McLaughlin & Marsh, 1990; Webster-Stratton, 1998). The aim of the developed program information was to provide some research-based knowledge about bully/victim problems and their characteristics and to give guidelines for how to deal with such problems. It is natural to regard this program information as a precondition to appropriate implementation, even though teachers may need other forms of support and assistance as well. In the present study, we measured the extent to which teachers had read the printed program information received, which was captured in the predictor Read Program Information.

    Values and attitudes. Green and Kreuter (1999, p. 40, pp. 161164) have emphasized the importance of teachers values and attitudes in their general model of program implementation. It is often argued that this point is particularly relevant with regard to schools because teachers have a high degree of professional autonomy and they typically perform their work in an isolated structure, within their own classrooms (Hargreaves, 1992; Kirby, Stringfield, Teddlie, & Wimpelberg, 1992; Lagerweij & Voogt, 1990; Nias, 1990. In the present study, we used two different indices of teachers attitudes toward bullying. One of these measures, Readiness to Intervene, focuses on the perceived importance of making or getting into contact with the parents of the implicated children in a possible bully/victim problem. The other measure, Information about Break Times, mainly reflects the perceived importance of having a good overview of the students activities during break periods.

    Affective involvement. We assume that teachers general emotional responsiveness and emphatic identification with victims of bullying will increase their motivation to counteract such problems. Two different measures were used in the present study: Affective Involvement and Self-Victimized as a Child.

    School factors. In the international literature on innovations in schools, it is usually claimed that the role of the principal/leadership group is vital to almost any successful changes in schools (Berman & McLaughlin, 1977, 1978; Fullan, 1991, 1997; Levine & Lezotte, 1990; Mortimore, 1988; Stoll, 1999). It has also been reported that other aspects of teachers work climate have considerable relevance for processes of change in schools. This point is clearly stated by Fullan (1991) when he concluded that the quality of working relationships among teachers is strongly related to implementation (p. 77; see also Fullan, 1998; McLaughlin, 1991, 1998; Stoll, 1999). In line with this research, our main interest was focused on variables reflecting aspects of the school climate and the schools general involvement with and attention to bullying problems. The school climate was measured with four previously derived dimensions (Kallestad, Olweus, & Alsaker, 1998): (a) TeacherLeadership Collaboration, (b) TeacherTeacher Collaboration, (c) Openness in

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Communication, and (d) Orientation to Change. In addition, we constructed an index labeled School Attention to Bullying Problems.

    Method

    ParticipantsThe present study is part of a large-scale, longitudinal research program under the direction of Dan Olweus, which is known as School as a Context for Social Development. The overriding focus of the project is on bully/victim problems in school and, in particular, on the implementation and effects of the Olweus Bullying Prevention Program (see, e.g., Olweus, 1991, 1992, 1993a, 1994, 1999b, 1999c; Olweus & Limber, 1999)

    Data covering a wide range of variables and reported by different groups of informants including students, teachers, principals, and parents were collected at several time points over a period of 2.5 years (May 1983October 1985). The key instruments of data collection were self-report questionnaires (see, e.g., Kallestad et al., 1998; Olweus, 1991).

    The sampling procedure was based on classes/schools in a quasi-experimental design with four different cohorts of students who were in Grades 69 in the presented analyses (according to the current grade system). Classes/schools were allocated to cohorts by a basically random procedure. For the present study, we analyzed teacher questionnaire data from two time points: here referred to as Time 1 (October/November 1983) and Time 2 (May/June 1984). All teachers included in the study were homeroom (main) teachers and, in accordance with Norwegian practice, taught a majority of the lessons in their own classrooms (on average, 19 lessons per week). The classes were fairly evenly spread over Grades 69. Only teachers with valid data at both time points were included in the multilevel analyses, which reduced the original, longitudinal sample from 112 to 89 teachers distributed across 37 schools. Data from the student participants were not included in the following analyses.

    The school system in Bergen (the second largest town in Norway, with approximately 200,000 inhabitants) does not deviate markedly from the school system in other town communities of Norway. Accordingly, the sample of teachers in the present study is likely to be broadly representative of teachers in the community of Bergen and other town communities of Norway as well.

    A Brief Description of the Olweus Bullying Prevention ProgramThe overriding goals of the program can be stated as follows (Olweus, 1993a): (a) to reduce as much as possible, if not eliminate, existing bully/victim problems inside and outside of the school setting; (b) to prevent the development of new bully/victim problems; and (c) to achieve better peer relations at school and create conditions that allow victims and bullies, in particular, to get along and function better inside and outside of the school setting.

    These goals may be summarized in one general statement to the effect that schools should be safe and

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • positive learning environments.

    Given the considerable stability of aggressive behavior over time (e.g., Olweus, 1979) and the generally low or modest success in reducing such behavior with a number of individual-oriented approaches, an important premise of the Olweus Bullying Prevention Program is that bullying behavior can be checked and redirected into more prosocial directions through a systematic restructuring of the social environment. Among other outcomes, this restructuring is expected to result in changes in the opportunity and reward structures for bullying behavior, resulting in fewer opportunities for bullying and fewer or smaller rewards (e.g., in the form of material rewards, prestige, or peer support) for displaying such behavior (see Olweus, 1993a, 2001b; Olweus & Limber, 1999).

    Generally, the Olweus Bullying Prevention Program is built around a limited set of key principles and findings derived chiefly from research on the development and modification of implicated problem behaviors; particularly, aggressive behavior (see, e.g., Olweus, 1993a, 1994). An overview of the program is presented in Table 1. More details about the specific measures to be used at the school, classroom, and individual levels are found in other publications (Olweus, 1993a, 2001b; Olweus & Limber, 1999).

    The intervention program was introduced to the schools in the fall (October/November) of 1983 and continued to be in place (to varying degrees) for the remaining period of the project. More concretely, the following materials were used in the intervention part of the project.

    A booklet for teachers/school personnel describing what is known about bully/victim problems (or rather, what was known in 1983) and giving guidelines and suggestions about what teachers and the school could do to counteract and prevent problems (Olweus & Roland, 1983). Efforts were also made to dispel common myths about the nature and causes of bully/victim problems that might interfere with an adequate handling of them. All main/homeroom teachers of the participating classes (and most other teachers in the participating schools) received a copy of this booklet.

    A four-page folder with information and advice for parents of victims and bullies as well as ordinary children. This folder was distributed by the teachers to the families of participating students.

    A 25-min video cassette showing episodes from the everyday lives of two bullied children: a 10-year-old boy and a 14-year-old girl. All participating schools received one or more copies of this video.

    As indicated previously, our small project group had regular contact with the schools in connection with the several surveys conducted. Fifteen months after the program was first offered to the schools, individual feedback information was presented in meetings with the staff at each of the 42 participating schools. These meetings focused on the level of problems in the particular school and the social environments (teachers, parents, and peers) reactions to the problems. The main principles of the program and the major procedures proposed for intervention were also presented and discussed. Generally, this addition to the program, as well as the program itself, was quite favorably received by the teachers, as expressed in their ratings and comments (Manger & Olweus, 1985).

    Two Measures of Implementation

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • The individual homeroom teacher is a key person in the implementation and maintenance of the Olweus Bullying Prevention Program. Accordingly, we developed two different indices of the degree of implementation at the teacher or classroom level: the Classroom Intervention Measures and Individual Contact indices (measuring contact with students and parents of involved students). Because we also wanted to explore possible school differences in degree of implementation, these two measures were aggregated at the school level as part of the statistical analyses.

    The Classroom Intervention Measures is an additive index consisting of seven specific intervention measures and three questions about the degree to which the individual teacher, in his or her own view, had involved him- or herself, the students in the class, and their parents in counteracting bullying during the past spring term (45 months in the Norwegian school system). The individual items of this index were scored on 2- or 3-point scales (01 or 02, respectively). Brief information on the content and coding of the items is presented in Table 2.

    Table 2Percentage of Teachers With a Positive Scorea on Items Included in

    Classroom Intervention Measures

    Items/scales %

    Classroom Intervention Measures (sum of items 110) 89Specific Intervention Measures (sum of items 17 below) 83

    1. Role-play 112. Literature 493. Class rulesb 274. Class meetingsc 255. Video 216. Information folder to parentsd 327. Other actions 23

    Involvement (sum of items 810 below) 648. General teacher involvemente 439. Involvement of students 53

    10. Involvement of parents 6

    Note. n (number of teachers) varying between 80 and 89. aApplied once or several times. bNo class rules against bullying (0) / had class rules already (1) / introduced class rules (2). cNo class meetings on bullying (0) / regular class meetings (2). dInformation folder distributed (1) / not distributed (0). eDegree of involvement recoded to 3-point scale (02).

    The correlation between the sum scores of the seven specific intervention measures and the three items on teacher involvement was .59 (p = .000). In addition, preliminary multilevel analyses with these two aspects of classroom intervention gave similar results. Accordingly, we decided to combine all 10 items of Table 2 into

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • one Classroom Intervention Measures index. However, to provide somewhat more detailed information, we also present summary data on these two different aspects of classroom intervention in Table 2 (named Specific Intervention Measures and Involvement, respectively, in the table).

    When teachers recognize or suspect that there is a bully/victim problem in their classroom, the program strongly recommends that they contact the students involved and, usually, their parents (cf. see Measures at the Individual Level in Table 1). Against this background, we constructed an index, Individual Contact, reflecting the degree of contact made with bullied/bullying students and their parents. There were thus four possible targets of contact, the bullied and the bullying student(s), respectively, and their parents. Each variable was scored dichotomously (1 = contact and 0 = no contact), and, accordingly, a teacher could get a maximum score of four on this index.

    Predictors in the Multilevel ModelsThe selected potential predictors were measured at either Time 1 or Time 2, with one exception in which data were available at both time points. Time of measurement (T1 and T2) is indicated within parentheses for each of the potential predictors in the following paragraphs.

    Perceived Level of Bullying (T1 + T2). Three items were used to measure the degree to which teachers perceived bullying as a problem in their own classrooms. The teachers reported the estimated number of students participating in bullying, being bullied, and being excluded by other students. The index was constructed as a mean score of the three items at Time 1 and Time 2. The internal consistency reliability of the scale (Cronbachs alpha) was .78.

    Perceived Staff Importance (T1). Three separate questions were used to measure the importance of the homeroom teacher, other teachers, and leadership group (including the principal and vice principal), respectively, with regard to counteracting bullying in school. Each question was answered on a 6-point scale, with responses ranging from 0 (not at all important) to 5 (very important). The empirical analyses showed that, on average, the teachers considered themselves, their colleagues, and the leadership group to be of almost equal importance for reducing bullying in school. The data showed considerable individual differences, however, and the index comprising the three items showed an internal consistency reliability of .80 (Cronbachs alpha).

    Read Program Information (T2). The teachers reported whether they had read the information booklet on bully/victim problems distributed to all teachers in the project. The response alternatives were 0 (no), 1(superficially), and 2 (yes). The teachers were also asked whether they had read the four-page folder developed for parents, with response categories 0 (no) and 1 (yes). The Read Program Information index was constructed as a sum score of the two items mentioned previously. The Pearson productmoment correlation between the two items was .32 (p = .003).

    Readiness to Intervene (T2). This indicates the extent to which teachers were inclined to make contact with the parents of a possible victim or bully in situations in which there was only a suspicion, and not clear evidence, of a bully/victim problem. Two items focused on the teachers own possible contacts with parents of victims and/or bullies, whereas the two remaining items asked about whether he or she thought such

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • contacts should be made by the parents in such somewhat ambiguous situations. The questions were scored on a 4-point scale, representing 0 (no), 1 (maybe), 2 (quite likely), and 3 (definitely yes). This index represents the mean score of the four items. The internal consistency reliability of the index (Cronbachs alpha) was .76.

    Information About Break Times (T1) The teachers reported whether they had a reasonably good overview of the students activities during break times and whether they considered such information to be important to them. These two items were scored on a 6-point scale, with responses ranging from 0 (not important at all) to 5 (very important). The index was constructed as a mean score of the two items. The Pearson correlation between the two items was .31 (p = .004).

    Affective Involvement (T2). Two items covered teachers attitude and affective involvement with regard to bullying. The teachers were asked about their emotional reactions to the individual students bullying of other students (reactions varying from [understanding] to [being very upset]), and what they felt if they saw or realized that a student was being bullied (from [feeling nothing] to [being very upset]). The items were scored on a 4-point scale (03). The index Affective Involvement was constructed as a mean score of the two items (Cronbachs alpha .68).

    Self-Victimized as a Child (T2). The teachers were asked about whether they themselves had been victimized as children in school. This single item was scored on a 4-point scale, with response categories indicating the extent to which they had been bullied (from [no], [never], to [yes, in several different time periods]).

    Teacher background variables (T1). Several background variables were included in the regression analyses: the teachers age and gender, number of years of education, years of service at the school, and type of employment contract (ordinary/substitute teacher). In addition, we included the grade level of the teachers homeroom class.

    School climate (T1). In considering potential, conceptually meaningful predictors at the school level, our main focus was on variables reflecting aspects of the school climate and the schools general involvement with and attention to bullying problems. As detailed in a previous paper (Kallestad et al., 1998), school climate was measured in this project (using the same samples of teachers as in this article) with four key, factorially derived dimensions: (a) TeacherLeadership Collaboration (e.g., The collaboration between the teachers and the leadership group is generally good), (b) TeacherTeacher Collaboration (Teachers at school are helpful toward each other), (c) Openness in Communication (I talk openly with the other teachers at school about my relationship with my students), and (d) Orientation to Change (I am very interested in trying new ways of dealing with students). These climate dimensions showed good psychometric properties (e.g., in terms of reliability and stability over time; see Kallestad et al., 1998) already at the level of the individual teacher; they had as expected even better such properties when the teacher data were aggregated by school.

    School Attention to Bullying Problems (T2). A special index, consisting of five items, measuring the schools attention to bullying problems was also constructed. The items referred to whether the school leadership group had presented results from the questionnaire survey (made in October/November 1983) for its own school to the staff and whether a special working group of teachers at the school had discussed and summarized the results. Included in this index were also questions about possible formal and informal staff discussions about bullying and whether the school had arranged a special school conference day on bullying.

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Three of the items were scored as yes or no (10), whereas two items were scored on a 3-point scale reflecting frequency of meetings and informal discussions (20). The responses to these questions derived from individual teachers, were aggregated across schools and combined into a sum score for each school.

    The Pearson correlations among the predictor variables at the teacher/classroom and school levels are presented in Appendix A and Appendix B, respectively.

    As indicated previously, the selected potential predictor variables were (with one exception) measured at either Time 1 (before implementation of the intervention program) or Time 2. Accordingly, when using the term prediction, we do it in the general regression-analytic sense of estimating values on an outcome variable on the basis of information from one or more other (predictor) variables. However, several of the relationships reported actually reflect true temporal prediction, with the predictor variables (such as Perceived Staff Importance and Information about Break Times) being measured about 6 months before the outcome variable. For other relationships, predictor and outcome variables were measured at the same time (Time 2; in some cases, however, the information contained in the predictor variable referred to events/activities that presumably occurred several months earlier, as with regard to Read Program Information, for example). The items included in School Attention to Bullying Problems were measured at Time 2, but most of these items referred to activities that normally took place in the early stages of the implementation process.

    Multilevel AnalysesAs mentioned, the data of the present study have a hierarchical structure with teachers nested within schools. Because of this structure, and with possible predictor variables at both the teacher and school levels, it is not only natural but actually required to analyze the results with multilevel model techniques (see, e.g., Bryk & Raudenbush, 1992; Goldstein, 1987).

    In the present study we used the ML3E-program (Prosser, Rasbash, & Goldstein, 1991) for the statistical analyses. Estimation in multilevel analyses is usually performed in three steps that are referred to in the next section as the unconditional model, the within-school model, and the between-schools model.

    Results

    Some Descriptive Data on the Implementation MeasuresAs shown in Table 2, 83% of the teachers reported that they had used in their own classroom at least one specific measure proposed in the Olweus Bullying Prevention Program. Use of literature (in order to define and make concrete the meaning of bullying and to increase empathy with victims) was the specific measure most frequently used (49%), whereas role-play was least frequently applied (11%). Table 2 also shows that 64% of the teachers reported at least one positive answer on the three-item Involvement index. It may be noted that few of the teachers (6%) had specifically involved parents in antibullying work at the classroom

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • level (in parent classroom meetings, for example; however, the percentage was clearly higher with regard to individual contacts, see next paragraph). Overall, a total of 89% of the teachers had at least one positive score on the summary Classroom Intervention Measures index.

    Table 3 shows that 60% of the teachers had at least one positive score (reported contact) on Individual Contact. The table also displays the percentage of teachers who had minimum one positive score on each of the individual items included in the summary index. Contact with students was more common than contact with parents, as expected. More detailed analyses showed that 58% of the teachers reported contact with students (positive score on Item 1 or Item 2 in Table 3), whereas 38% of the teachers had made contact with parents of involved students (positive score on Item 3 or Item 4 in Table 3).

    Table 3Percentage of Teachers With a Positive Scorea on Items

    Included in Individual Contact

    Items/scales %

    Individual Contact (sum of items 14 below) 601. Contact with victimized student 402. Contact with bullying student 443. Contact with parents of victimized student 244. Contact with parents of bullying student 25

    Note. n (number of teachers) = 89.aAt least one contact in the past 34 months.

    Decomposition of Variance in Outcome Measures (Unconditional Model)First, a simple variance components model was fitted to the two outcome variablesClassroom Intervention Measures and Individual Contactto decompose the variance into a within-school and a between-school component. In this model, the unconditional model, no predictor variables were included. Estimation of the within-school (2) and the between-school variances () permits calculation of the total variance (2 + ) for the outcome measures and the proportion of the total variance in individual teacher reports that is within and between schools, respectively.

    The between-school variance was estimated at .14 for Classroom Intervention Measures and at zero for Individual Contact. There was thus no reliable between-school variance for the Individual Contact index, whereas 14% of the total variance in the Classroom Intervention Measure index was reliable variance between schools (cf. Kallestad et al., 1998). Introduction of within-school (teacher-level) predictors will usually reduce the within-school variance. It should be noted, however, that introduction of within-school predictors may also affect the estimate of the between-schools variance, making it smaller or larger depending on the pattern of relations among the predictors and the outcome variable (see next section).

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Predicting Implementation With Teacher-Level Predictors: The Within-School ModelsIn the second stage of analysis, we predicted the two implementation measures on the basis of the selected predictors at the teacher/classroom level. Estimation of the within-school model was performed in the following way, modeled on the approach used by Rowan, Raudenbush, and Kang (1991). In the first run, we entered all predictors into the model. Predictors with regression coefficients less than 1.5 times their estimated standard error were dropped from the analyses, and the model was reestimated. In the final model, the coefficients had to be at least twice their standard errors to be considered significant (but near-significant predictors were kept in the model). Proportion of predicted variance was calculated by subtracting the new (residual) within-school variance component from the corresponding component in the unconditional model (without predictors) and by dividing by the latter component.

    Predictors and proportions of predicted variance are reported in Table 4. Although for convenience the table also shows the results for the between-school model, in this section we focus on the within-school analyses.

    Table 4Predicting Classroom Intervention Measures

    PredictorRegressioncoefficient

    Standarderror

    (SE)

    Within-school predictorPerceived Staff Importance 1.552** .288Read Program Information 1.272** .280Perceived Level of Bullying 1.310* .465Self-Victimized as a Child 0.724* .332Affective Involvement 0.609* .299

    Between-school predictorOpenness in Communication 1.270* .519School Attention to Bullying Problems 0.517* .228TeacherTeacher Collaboration 1.100 .770

    Unconditional varianceWithin schools 9.055Between schools 1.528

    Residual variance (within model)Within schools 4.216Between schools 2.678

    Residual variance (between model)

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Within schools 4.322Between schools 1.335

    Percentage of explained variance Within schools 53.4a

    Between schools 50.2b

    Note. n = 84.a Percentage variance explained is calculated as the difference in variance estimates between

    unconditional model and within-school model: 53.4%: (9.055 4.216) 100/9.055. b Percentage variance explained is calculated as the difference in between-school variance estimates in within-school model (residual variance) and between-school model: 50.2%: (2.678 1.335) 100/2.678.* Denotes a coefficient (an unstandardized regression coefficient) that is at least twice as large as its standard error (SE). ** Denotes a coefficient that is at least three times as large as its standard error.

    As evident from Table 4, 53.4% of the within-school variance (a reduction from 9.055 to 4.216) in degree of implementation of the Classroom Intervention Measures could be predicted by the five predictors listed in the upper panel of the table (Perceived Staff Importance, Read Program Information, Perceived Level of Bullying, Self-Victimized as a Child, and Affective Involvement).

    One of the two strongest predictors was Read Program Information indicating that teachers who had themselves actually read the distributed information about the program, the teacher booklet, and the four-page parent folder implemented more of the Classroom Intervention Measures. Similarly, teachers who believed that the actions of the staff (themselves, other teachers, and the school management) were important for the outcome in counteracting bullying in the classroom (Perceived Staff Importance) implemented more of the specific classroom measures.

    Furthermore, teachers who perceived more bully/victim problems in their own classrooms, as indicated by Perceived Level of Bullying, tended to have higher scores on the Classroom Intervention Measures. The final two significant predictors were both related to the individual teachers degree of affective involvement in bully/victim problems. Teachers who reported that they had themselves been exposed to some degree of victimization in their school years (Self-Victimized as a Child) and who reacted more negatively and were more upset about victimization of students (Affective Involvement) tended to implement more of the Classroom Intervention Measures.

    It should be noted that all of the teacher background variables (Appendix A) were only marginally related to the implementation measure and, therefore, were not included in the final model.

    The results from the within-school model, with regression coefficients standardized (the B coefficients multiplied by the ratio of the within-school standard deviation in the relevant predictor variable to the within-school standard deviation in Classroom Intervention Measures), are graphically displayed in Figure 1. It may be noted that the correlations among the predictors were generally quite low, the highest value being .23 (p = .031, between Self-Victimized as a Child and Perceived Level of Bullying; Appendix A).

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Figure 1. Teacher-level factors predicting implementation of Classroom Intervention Measures (within-school model). Figures close to arrows are standardized regression coefficients.

    With regard to Individual Contact, 33.6% of the within-school variance (a reduction from 1.856 to 1.233) could be predicted by the five predictors listed in Table 5. Two of these predictors, Perceived Level of Bullying and Perceived Staff Importance, were also significant in predicting Classroom Intervention Measures. The third significant predictor was Information about Break Times (see Table 5 and Figure 2).

    Table 5Predicting Individual Contact

    PredictorRegressioncoefficient

    Standarderror

    (SE)

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Within-school predictorPerceived Level of Bullying 1.063** .209Perceived Staff Importance .289* .139Information About Break Times .320* .155Readiness to Intervene .296 .190Ordinary/Substitute Teacher .710 .374

    Unconditional varianceWithin schools 1.856Between schools 0

    Residual varianceWithin schools 1.233Between schools 0

    Percentage of explained variance Within schools 33.6a

    Between schools 0

    Note. n = 89.a Percentage variance explained is calculated as the difference in variance estimates

    between the unconditional model and within-school model: 33.6%: (1.856 1.233) 100/1.856.* Denotes a coefficient (unstandardized regression coefficient) that is at least twice as large as its standard error (SE). ** Denotes a coefficient that is at least three times as large as its standard error.

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Figure 2. Teacher-level factors predicting degree of Individual Contact with involved students/their parents (within-school model). Figures close to arrows are standardized regression coefficients.

    In the final model for Individual Contact there were also two near-significant predictors: Readiness to Intervene and Substitute/Ordinary Teacher. As expected, the contact readiness variable was positively related to Individual Contact. In addition, teachers with a permanent employment contract, in contrast to substitute teachers, tended to have higher scores on Individual Contact. The other teacher background variables were again only weakly related to implementation and, consequently, were not included as predictors in the final model. The correlations among the predictors were generally quite weak also in this model, with a highest value of .27 (p = .011; between Perceived Staff Importance and Readiness to Intervene; Appendix A).

    Predicting Implementation with School-Level Predictors: The Between-Schools ModelAs evident from Table 5, there was no reliable between-school variance for Individual Contact in the Variance Components Model (Unconditional Model), and this estimate did not change after the teacher-level predictors were included in the analyses. Accordingly, it was not considered meaningful to pursue the

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • analyses at the school level for this implementation variable.

    In contrast, for the Classroom Intervention Measures, the between-school variance became larger (an increase from 1.528 to 2.678; see Table 4) when predictors at the teacher-level were entered and controlled for. This implies that the scores on the within-school predictors were not evenly distributed across schools. When this fact was taken into account and adjusted for in the within-school model, the between-school variance in implementation actually increased (the residual between-school variance now constituting 39% of the total residual variance). This result indicates that prediction of the Classroom Intervention Measures could also be meaningfully pursued at the school level. For these analyses, it is natural to use the adjusted (residual) between-school variance as a baseline for calculating proportion of predicted variance. Generally, estimation of the between-schools models followed the same procedure as those described for the within-school models.

    Table 4 shows that 50.2% of the between-school variance could be predicted by three school-level predictors (the school climate dimensions): Openness in Communication, TeacherTeacher Collaboration (nonsignificant), and the index School Attention to Bullying Problems. If only the two significant predictors were included in the model, 43.5% of the variance could be predicted.

    As suggested previously, the items making up the school climate dimension Openness in Communication mainly concern questions about frequency of informal conversations with other colleagues about classroom work and relationships with students (see Kallestad et al., 1998). Schools characterized by openness in communication among staff were thus more likely to implement more of the Classroom Intervention Measures.

    It is also worth noting that the other school climate dimension in the table, TeacherTeacher Collaboration, somewhat surprisingly, showed a negative relationship with the outcome variable (though it did not quite reach significance in the final analyses). The implication is that schools characterized by good collaboration among the teachers actually tended to implement less of the intervention program.

    The special index School Attention to Bullying Problems, described previously, was also clearly related to the schools degree of implementation of the Classroom Intervention Measures. Teachers in schools that had, for example, presented the questionnaire survey results in a plenary meeting and arranged a special school conference day on bullying were thus more likely to implement the specific Classroom Intervention Measures.

    As shown in Appendix B, two pairs of variables were substantially correlated in school-level analyses, the TeacherTeacher Collaboration and the TeacherLeadership Collaboration dimensions, on one hand (r = .68, p = .000; Appendix B), and the Openness in Communication and Orientation to Change dimensions, on the other (r = .52, p = .001; Appendix B). This is likely to have made it more difficult to disentangle the unique contributions of the individual variables in these variable pairs in the school-level analyses.

    Discussion of ResultsThe present study shows that at least some part of the Olweus Bullying Prevention Program was actually put

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • into practice by most teachers in our sample. Almost 90% of the teachers had at least one positive score on Classroom Intervention Measures, and as much as 83% had at least one positive score on the items making up the subcategory Specific Intervention Measures in Table 2. Although this may be seen as a relatively satisfactory overall result, it should be noted that the degree of implementation varied considerably among teachers (and schools). In particular, more advanced or maybe demanding measures such as role-play, class rules, and class meetings were used less frequently.

    To foreshadow results to be reported in a separate article, degree of implementation of a particular intervention component did not correspond directly with its effectiveness in reducing bully/victim problems. Nonetheless, there was overall a significant and substantial correlation between degree of implementation of the Classroom Intervention Measures as a general index and degree of reduction of bully/victim problems.

    A positive score on Individual Contact presupposes, as mentioned earlier, that the teacher had identified, or suspected the existence of, a bully/victim problem in his or her classroom. As evident from Table 3, 60% of the teachers had a positive score on Individual Contact. As long as we do not know how many of the teachers remained passive after having identified a (possible) bullying problem in their own classrooms, it is difficult to give this result a precise interpretation. Nonetheless, 60% seems to be a relatively high percentage in consideration of the fact that the reference period for reporting was this springa period of 45 months.

    With regard to the key aim of the present study, that of predicting degree of implementation of the intervention program, the results of the empirical analyses must be considered successful. For both outcome variables, Classroom Intervention Measures and Individual Contact, substantial amounts of variance, 53.4% and 33.6%, respectively, were accounted for by the teacher-level predictors in the final within-school models; for the Classroom Intervention Measures, a considerable 50.2% of the (adjusted) between-school variance could be predicted by three school-level predictors. As reported in the Results section, there was no reliable between-school variance for the Individual Contact index, and, accordingly, no school-level models were estimated for this outcome variable.

    We now continue with a more detailed discussion of the empirical findings, starting with the within-school models followed by the between-school model for Classroom Intervention Measures.

    Teacher-Level PredictorsPerceived Level of Bullying (in own class) was important with regard to prediction of both outcome variables: Individual Contact (with a standardized beta coefficient of .45; see Figure 2) and Classroom Intervention Measures (with a standardized beta coefficient of .25; see Figure 1). It is natural to expect that teachers who have recognized, or suspect the existence of, bully/victim problems in their own classrooms will be more likely to make contact with involved students and their parents. It is also quite understandable that the teachers perceptions of the level of bullying problems are likely to affect their willingness to introduce more specific classroom measures.

    As seen in the uppermost part of the program overview in Table 1, awareness and involvement on the part of adults are seen as important general prerequisites to effective implementation of the program. In our recent intervention work, systematic use of the Olweus Bully/Victim Questionnaire (Olweus, 1996, in press; Solberg

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • & Olweus, 2003) is considered an important vehicle for raising awareness and thereby increasing involvement of the adults at school, in particular. To assess the level of problems, we strongly recommend schools to carry out a survey with the questionnaire in an early phase as part of the intervention package.1 This is likely to give the teachers a reasonably realistic picture of the situation with regard to bully/victim problems in their own school and thereby increase their readiness to engage in antibullying work, when needed (which is actually needed in most schools, as of today). We also recommend schools to administer the questionnaire 1 year after the first assessment to find out what may have been achieved or not achieved and what needs continuing or newly implemented efforts.

    It is important to emphasize, however, that perceived level of bullying problems was by no means the only important predictor of the outcome (implementation) variables concerned. With regard to Classroom Intervention Measures, this index was most strongly predicted by Perceived Staff Importance (with a standardized beta coefficient of .47; see Figure 1). This result implies that teachers who saw themselves, their colleagues, and the schools as important agents of change for counteracting bully/victim problems among their students were more likely to involve themselves in antibullying efforts and to introduce specific classroom measures. This finding is clearly in line with previous research concerning teacher efficacy, which was mentioned in the introduction of the article. It is natural to assume that this result also implies a belief on the part of these teachers that it was actually possible to reduce the level of such problems through systematic classroom and school activities.

    Perceived Staff Importance also predicted Individual Contact, but somewhat less strongly than Classroom Intervention Measures (with a standardized beta coefficient of .19). This may possibly indicate that teachers considered such actions more demanding or difficult and that they were less convinced about the success of such contacts.

    It may be suggested that Perceived Staff Importance not only measures teachers perceived influence but very likely also reflects their perceived responsibility for doing something about bullying in their own classrooms. With such an interpretation, this result also implies that teachers who saw it as their own, as well as the schools, task and professional responsibility to counteract bully/victim problems among their students were more likely to involve themselves in antibullying efforts and to introduce specific classroom measures. Forceful legislation in this area, with responsibility for preventing and counteracting bullying problems being clearly placed with the teachers and the school leadership, is certainly one of several worthwhile means of increasing teachers perceived importance and responsibility on this point (cf., e.g., Olweus, 1999b; Smith, 1999).

    The remaining predictors in Tables 4 and 5 were not common for the two outcome measures. Another important predictor of Classroom Intervention Measures was the variable Read Program Information (with a standardized beta coefficient of .36; see Figure 1). The information materials, the teacher booklet, and the parent folder had two key aims: to provide some research-based knowledge about bully/victim problems and to give guidelines for how to deal with such problems. A straightforward interpretation of this result is that teachers who read more of the program information became more motivated to do something about the problem and, in particular, acquired more knowledge about how to counteract the problems through suggested classroom activities. This perspective was suggested in the introduction. It is, however, also possible that teachers who were highly involved and motivated at the outset read more of the program materials and were likely to implement proposed classroom activities to a greater extent than less motivated colleagues. If the second interpretation is also true (for some teachers), the possible implicated mechanisms

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • are likely to be more complex and may involve reciprocal causal processes between the Read Program Information and the unmeasured motivation variable. Nonetheless, and irrespective of the underlying mechanisms, having read the available program materials and acquired knowledge about the various program components can be seen as an important prerequisite to implementation of the program.

    The information about the program and its components provided to teachers can in principle be made more or less stringent and directive with regard to guidelines and other requirements. Additionally, as has been pointed out in the literature, teachers general attitude to experts and highly structured programs seems to vary across different time periods (Cochran-Smith & Lytle, 1999; Fullan, 1982, 1998). The relative lack of detailed advice and how-to-do-it information was probably an advantage for the project at the time when it was initiated. The Zeitgeist has changed markedly over the past 1520 years, however. Today, teachers ask for more detailed instructions, guidelines, and ready-made work materials, as also reflected in a new teacher manual (for details, see Olweus, 2001b; Olweus & Limber, 1999). Also, in order to secure better understanding and knowledge of the program and its implementation in classroom and break periods, we have usually established staff discussion groups at each intervention school, under the leadership of persons who have received special training about the program. All of these activities and other quality control measures are likely to enhance program fidelity, resulting in more uniform implementation of the program. This will also imply less between-teachers and between-schools differences in implementation, which is probably considered a desirable goal by most designers of intervention programs for our time.

    The two remaining significant predictors of Classroom Intervention Measures were Affective Involvement and Self-Victimized as a Child. These variables concern the teachers general emotional responsiveness and empathic identification with victims of bullying. Although the relationship of these two variables with Classroom Intervention Measures was somewhat weaker than was the case with the three predictors discussed earlier, the results point to the importance of the teachers affective involvement for a successful result, here defined as a relatively high degree of implementation of the program.

    This finding may have some implications for teachers who reported weak emotional reactions to bullying among students. Judging from reports by parents, it seems reasonable to believe that at least some teachers see victims of bullying in a fairly negative light, as a kind of nuisance that only creates problems and extra work for them. It is of course important to try to change such detached or even hostile views of victims, if present. Research results on the typical characteristics of victims (Olweus, 1993a) may be of some help in giving teachers a more realistic picture of the degrading and distressing situation of most victims, as well as providing long-term consequences of persistent victimization (e.g., Olweus, 1993b). Perhaps even more important would be the use of detailed real-life, literary case descriptions or a well-designed video (such as the Norwegian Bullying: Scenes from the Everyday Lives of Two Bullied Children from 1983 or its American counterpart Bullying from 1996) that may increase empathic responding and help teachers (and students) see the situation at least, in part, from the perspective of the victim. As suggested by our empirical analyses, such affective involvement is likely to increase the teachers readiness to counteract bullying in their own classrooms.

    With regard to the other outcome variable, Individual Contact, there was one more significant predictor in the final model (Table 5): Information about Break Times. Also, this result makes good sense, indicating that awareness about what goes on among the students during break periods is an important prerequisite to active intervention with (possible) victims and bullies as well as with their parents.

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Two more predictors of Individual Contact were retained in the model (Table 5), although they did not quite reach significance in the final analyses, Readiness to Intervene and Substitute/Ordinary Teacher. The first of these variables reflects the teachers willingness to be contacted by the parents, or to make contact themselves with the parents, if there is only a suspicion that a student is involved in bully/victim problems. Teachers with a low threshold for such contacts were somewhat more likely to initiate contact with bullied or bullying students and their parents than teachers with a higher threshold.

    Finally, teachers with a temporary employment contract (substitute teachers) made fewer contacts than teachers on the permanent staff (ordinary teachers). This result may be interpreted to suggest that the substitute teachers were less knowledgeable of the peer relationships among the students in the class, making it more difficult to identify bully/victim problems among the students. In addition, or alternatively, these teachers who are often newcomers to the school might be more reluctant to address these issues because of the possible trouble or negative attention such action may entail.

    Before closing this section, we call attention to the lack of relationship between degree of implementation and the age variable. The common assumption that younger teachers have a more positive attitude to change and, consequently, are more open to new ideas and programs than their older colleagues received little support in our study (cf. Huberman, 1988, p. 129). More generally, background variables such as teachers age, gender, and professional experience seemed to be of little importance with regard to adoption and use of the intervention program.

    School-Level PredictorsFirst, we want to reiterate the finding that teachers degree of implementation of Classroom Intervention Measures varied systematically among schools, whereas degree of Individual Contact did not. This latter result suggests that few or no schools in the study had a clear school policy about handling identified or suspected bully/victim situations; it was largely up to the individual teacher to decide how to address such problems. On the basis of this result, we limited our efforts to identify significant and meaningful predictors of the Individual Contact variable to the teacher/classroom level. For Classroom Intervention Measures, however, we also examined potential predictors at the school level.

    Our school-level analyses focused on work-related aspects of the school climate and the schools attention to bully/victim problems. A major finding was that schools with a higher degree of Openness in Communication among the teachers implemented more of the Classroom Intervention Measures (with a standardized beta coefficient of .49). At the same time, considering the zero-order correlations, it is obvious that the climate variable Orientation to Change was also substantially related to the Classroom Intervention Measures variable (r = .48, p = .003; see Appendix B). However, much of the variance predicted by Orientation to Change was also predicted by the Openness in Communication variable, thereby reducing the contribution of the Orientation to Change variable to a nonsignificant level in the between-school model. Because of the sizeable correlation between these two school climate predictors (r = .52, p = .001; see Appendix B), it seems reasonable to extend somewhat the previous conclusion to the effect that schools characterized by openness in communication among staff and a generally positive attitude to change were particularly likely to implement the program.

    Our study thus shows that school climate measures were important predictors of school differences in

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • teachers implementation of the program. This finding relates to Fullans (1991) statement, mentioned in the introduction, that the quality of working relationships among teachers is strongly related to implementation (p. 77). If we define quality of working relationships as openness in communication and positive attitude to change, our study clearly supports this statement. From this perspective, it will probably be easier for a school to meet new challenges and, for example, implement a new school-based intervention program if the school has already acquired a basic capacity to handle changes. Our study suggests that openness in collegial communication and a generally positive attitude to change among staff may be important resources in such processes.

    However, if we define quality of working relationships by reference to the TeacherTeacher Collaboration variable, our results actually run counter to Fullans (1991) generalization and our expectations. In the between-school model in Table 4, there was a negative relationship between this variable and Classroom Intervention Measures (the regression coefficient was nonsignificant, but the zero-order correlation was rather substantial, r = .34, p = .037; see Appendix B). Obviously, high general satisfaction with the working relationships with other teachers does not always promote a proactive attitude toward problems and new challenges. It may even be the other way around as the negative relationship suggests, implying that teachers actually try to avoid changes in schools where the collegial collaboration is particularly good. Such an interpretation is strengthened by the substantial negative correlation of .40 (p = .014; see Appendix B) between TeacherTeacher Collaboration and (positive) Orientation to Change. Although somewhat unexpected, these results are actually consistent with statements made by educational researchers such as Little (1990) who said that to promote increased teacher-to-teacher contact may be to intensify norms unfavorable to children (p. 524). At the very least, our findings do not support the common view that improvement of the situation for the students in a school must start by creating better quality of the working relationships among teachers.

    The negative relationship between TeacherTeacher Collaboration and degree of classroom implementation can also be assessed in light of the extensive research conducted by Andy Hargreaves (1992, 1994). Hargreaves described several typical school cultures, one of them being labeled Individualism. This type of school culture is seen as generally resistant to a range of various changes, including implementation of externally imposed programs or innovations. Schools characterized by Individualism tend to have staff rooms with a pleasant atmosphere and very little disagreement among the teachers. The negative relationship in our model may thus suggest that schools with good TeacherTeacher Collaboration and little program implementation were characterized by an individualistic culture in Hargreavess (1992, pp. 221233) sense.

    At the same time, it should be emphasized that the TeacherTeacher Collaboration variable is a reflection of the teachers general satisfaction with the collegial relationships and the degree of staff consensus about working and teaching methods at the particular school; it does not reflect in detail what the teachers actually agree on. It may be an important task in future studies to examine more precisely the contents of such collegial consensus and to explore under what conditions various content structures have positive or negative effects, or no effects at all.

    On the basis of earlier research, we expected to find a relationship between principal leadership and implementation. Against this background, we were somewhat surprised to find that TeacherLeadership Collaboration did not predict implementation of the program. It must be admitted, however, that this climate variable was also based on items formulated in fairly general terms. Accordingly, we had very little

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • information in this study about the degree of support or relative lack of support of the program provided by the principal/leadership group. It is reasonable to believe that the role of the principal/leadership group would have been more prominent in our prediction models if more specific information had been available and included among the predictors. At the same time, our results, with relatively limited between-schools variance, suggest that the teachers could largely decide on their own to what extent they wanted to implement components of the program in their own classrooms, irrespective of the principals permission or direct support. Such a view of the teachers as highly independent and in strong control of the organization of their own classroom work is clearly consistent with the teacher self-report data presented in our earlier article (Kallestad et al., 1998).

    The third school predictor in the between-school model in Table 4 was School Attention to Bullying Problems (with a standardized beta weight of .41; see Figure 3). The most reasonable interpretation of this result is that schools with more bullying-related activities for all or most of the staff generated an interest in the program and increased motivation among the schools teachers to implement key program components. This interpretation is based on the likely assumption that most of the activities included in this index took place in the early stages of the implementation process.

    On the other hand, it is also possible to reason that schools with a higher score on School Attention to Bullying Problems were more motivated to address the problems already at the outset, resulting in both the arrangement of schoolwide activities and greater implementation of specific classroom intervention measures. The results can thus be interpreted according to different causal-analytic models, and the data available do not permit a clear-cut choice between these models. If we choose a conservative strategy, excluding School Attention to Bullying Problems from the predictors, slightly more than a quarter of the between-school variance in Classroom Intervention Measures, 27.2%, can be predicted.

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • Figure 3. School-level factors predicting implementation of (aggregated) Classroom Intervention Measures (between-school model). Figures close to arrows are standardized regression coefficients.

    Relating to the previous discussion of the role of the principal, it is also possible to see the positive relationship between School Attention to Bullying Problems and the outcome variable as an indirect indication of the principals influence on teacher implementation of the intervention program. It is easy to see that a principal may influence staff attitudes and behavior by putting antibullying work on the schools official agenda, by initiating plenary meetings with staff and parents, and by providing clear guidelines about the organization of the supervisory system during break periods, for example. For antibullying work to get official recognition, it is also important that the principal allocates time and financial resources to such activities (cf. Green & Kreuter, 1999, pp. 191199), in addition to giving psychological support.

    In summary, the results from our study indicate that the teachers were without doubt the key agents of change with regard to adoption and implementation of the Olweus Bullying Prevention Program in school. Substantial amounts of variance in implementation could be predicted on the basis of our teacher- and school-level variables. Generally, we think our study has shed light on several factors of importance and has contributed to a better understanding of the process of program implementation. The empirical results have also suggested ways in which implementation of the program can be improved, and several of these amendments have already been incorporated in the program and its dissemination (Olweus, in press).

    References Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs,

    This

    doc

    umen

    t is c

    opyr

    ight

    ed b

    y th

    e A

    mer

    ican

    Psy

    chol

    ogic

    al A

    ssoc

    iatio

    n or

    one

    of i

    ts a

    llied

    pub

    lishe

    rs.

    This

    arti

    cle

    is in

    tend

    ed so

    lely

    for t

    he p

    erso

    nal u

    se o

    f the

    indi

    vidu

    al u

    ser a

    nd is

    not

    to b

    e di

    ssem

    inat

    ed b

    road

    ly.

  • NJ: Prentice Hall.

    Berman, P., & McLaughlin, M. (1977). Federal programs supporting educational change: Vol. VII. Factors affecting implementation and continuation. Santa Monica, CA: Rand Corp.

    Berman, P., & McLaughlin, M. (1978). Federal programs supporting educational change: Vol. VIII. Implementing and sustaining innovations. Santa Monica, CA: Rand Corp.

    Biglan, A. (1995). Changing cultural practices: A contextualist framework for intervention research. Reno, NV: Context Press.

    Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: Applications and data analysis methods. London: Sage Publications.

    Calhoun, E., & Joyce, B. (1998). Inside-out and outside-in: Learning from past and present school improvement paradigms. In A. Hargreaves, A. Lieberman, M. Fullan, & D. Hopkins (Eds.), International handbook of educational change (pp. 12861299). London: Kluwer Academic.

    Cochran-Smith, M., & Lytle, S. L. (1999). The teacher research movement: A decade later. Educational Researcher, 28(7), 1525.

    Fullan, M. (1982). The meaning of educational change. New York: Teachers College Press.

    Fullan, M. (1991). The new meaning of educational change. London: Casell.

    Fullan, M. (1997). Leadership for change. In M. Fullan (Ed.), The challenge of school change (pp. 115136). Arlington Heights, IL: Iri/Skylight.

    Fullan, M. (1998). The meaning of educational change: A quarter of a century of learning. In A. Hargreaves, A. Lieberman, M. Fullan, & D. Hopkins (Eds.), International handbook of educational change (pp. 214231). London: Kluwer Academic.

    Ghaith, G., & Yaghi, H. (1997). Relationships among experience, teacher efficacy, and attitudes toward the implementation of instructional innovation. Teaching and Teacher Education, 13(4), 451458.

    Goldstein, H. I. (1987). Multilevel models in educational and social research. London: Oxford University Press.

    Green, L. W., & Kreuter, M. W. (1999). Health promotion planning: An educational and ecolo