Improving Student Social Behavior: The Effectiveness of a Storytelling-Based Character Education Program

LifeStories for Kids uses the ancient art form of storytelling to make complex social skills, character choices, and social dilemmas more concrete and recognizable for children (Brightwood & DeRosier, 2007). The medium of storytelling is used to capture children’s attention and imagination while teaching positive social skills and character through identification with the story characters and situations. Accordingly, the incorporation of storytelling techniques has improved student ethical understanding in a CE program (Leming, 2000) and reduced student aggressive behavior in a SST program (Teglasi & Rothman, 2001). Storytelling techniques, however, have not been applied to a combined CE and SST approach prior to the development of the LifeStories for Kids curricula. Furthermore, efforts have not been made to maximize the impact of the storytelling medium in CE and SST programs through the use of live-action storytellers, multimodal presentation formats, and original artwork as visual cues. LifeStories for Kids uses these storytelling innovations to increase student attention, motivation to participate, and retention of story lessons.

Stories, read or told, have always been a primary teaching tool of educators (Lickona, 1991). Teachers use stories for a number of reasons: stories and storytelling capture children’s attention, encourage imagination, and enhance conceptual learning. Stories invite participation by guiding children in the creation of mental imagery and allowing them to experience the scene and its consequences through imagination (Coles, 1997; Norfolk & Norfolk, 1999). Research has shown that storytelling is a powerful teaching tool that enhances children’s memory of the story’s lesson as well as later learning and performance (Ahsen, 1996; Lickona, 1991). Stories also stir emotions, which neuroscience research suggests is a particularly effective tool for improving retention and memory (Caine & Caine, 1994). Because children are engaged cognitively and emotionally, learning becomes an active, experiential process with storytelling, rather than passive, didactic instruction.

Storytelling is particularly useful for teaching complex CE traits because storytelling teaches by illustration, making abstract concepts concrete and accessible (Vessels, 1998). For example, the use of storytelling has been supported as an effective strategy for moral education (Bebeau, 1993; Vitz, 1990) and a useful way to communicate and clarify values, particularly with young children (Bennett, 1995). Examples provided through narrative and visual images are an effective way of teaching and influencing behavior, particularly because they provide a meaningful context within which to understand the information (Coles, 1997). In contrast, didactic lecturing on virtues is not likely to be effective in teaching values to children due to the single-sided nature of the interaction. Kohlberg has argued that cognitive-developmental constraints preclude didactic teaching of values with children, and moral education is better fostered through dialogue and guided experience in moral reasoning (Kohlberg, 1985, 1983). Scholars agree that this kind of active participation through imagination and mutual storytelling engenders reasoning skills and the generalizability of concepts (Coles, 1997; R. A. Gardner, 1971).

Storytelling, when applied to CE, can involve children reading books or oral presentation of stories. When stories are orally presented, the power of the story to teach concepts is maximized (R. A. Gardner, 1971; Olmedo, 1997; Stucky, 1995), and the more engaging, animated, and demonstrative the storyteller, the more powerful the storytelling experience (Norfolk & Norfolk, 1999). Indeed, complex folktales, fables, and historical tales are the preferred medium for moral education due to their use of engaging characters and trials and errors to teach moral lessons (Stallings, 1997). Consequently, storytelling—specifically the creation of captivating stories targeting specific character traits as told by professional storytellers—was the method of presentation used for the development of the LifeStories for Kids series.

A comprehensive intervention curriculum is built around the stories included in the LifeStories for Kids program. Each story is accompanied by a set of classroom-based activities to extend the lesson for each story and provide multiple opportunities to practice new skills in structured social learning and academic exercises. The intervention developers drew on scientifically supported SST methods [e.g., social skills group intervention (DeRosier, 2004, 2007); social responsibility training (Guerra, Tolan, Eargle, Mosley, & Dunn, 1993), coaching (Oden & Asher, 1977), and anger coping (Lochman, Coie, Underwood, & Terry, 1993)] to create the classroom lessons, including didactic instruction with group process (e.g., discussion), active practice (e.g., games, directed exercises), modeling, role playing, positive reinforcement, and cognitive reframing.

Each lesson provides multiple structured activities from which teachers may choose. The inclusion of multiple activity options for each story permits teachers to reinforce learning through repeated exposure and step-wise instruction (e.g., moving from external to internal reflection), both of which have been shown to increase learning and memory for instructional material (e.g., Boud, Keogh, & Walker, 1985). In addition, research supports that students learn through multiple channels (e.g., visual, oral, tactile, experiential) and that educational efforts employing multiple modes of teaching are more effective for learning and memory than ones that do not (H. Gardner, 2004; Vincent & Ross, 2001). The variety of activities available in the LifeStories for Kids program enables teachers to teach skills and concepts through multiple modes of learning. Multi-modal education is also a more inclusive strategy reaching a broader range of students with different learning styles. The varied developmental levels of students are accommodated through the variety of multimodal activities available, as well as through separate developmental versions of activities within the LifeStories for Kids program.

LifeStories for Kids includes two parallel, but separate curricula, one for kindergarten through second grades (K-2) and another for third through fifth grades (3-5).

The nine stories (lasting 5-10 minutes each) in the lower elementary curriculum are told by nationally acclaimed, professional storyteller Willy Claflin (see http://www.willyclaflin.com). Though multiple character traits are addressed within each story, nine primary character traits are targeted: courage, good judgment, integrity, kindness, perseverance, respect, responsibility, self-discipline, and togetherness. Mr. Claflin’s telling of each story is presented via video with original, illustrative artwork integrated into the video as well, as throughout the curriculum materials, to serve as visual cues for story content. Following presentation of the videotaped story, teacher-led group discussion engages students in defining and expounding on the story’s character traits and lessons. Classroom and homework activities for skill reinforcement and practice are incorporated based on allotted time and teacher choice. For the K-2 curriculum, teachers may choose from the following set of activities: (a) flipbooks for group retelling with process discussion activities; (b) creative storytelling for students to extend the story; (c) role plays for students to explore the meaning and consequences of the story lesson; (d) art activities and coloring pages for students to creatively explore story content and relate story lessons to personal experience; and (e) activity worksheets to link story lessons with other academic areas. The combined activities last approximately 60 to 90 minutes per story. It is recommended that story presentation, discussion, and classroom activities be presented on multiple school days across a 1- to 2-week time span.

The nine stories (lasting 5-10 minutes each) in the upper elementary curriculum are told by three nationally acclaimed, professional storytellers: Willy Claflin (http://www.willyclaflin.com), Olga Loya (http://www.olgaloya.com), and Donna Washington (http://www.dlwstoryteller.com). The stories target multiple social dilemmas and character traits, specifically (grouped by story): (a) benefits of mutual respect and consequences of selfishness; (b) benefits of cooperation and value of diversity; (c) compromise and teamwork; (d) perspective taking and honesty; (e) assumptions based on differences and fairness; (f) courage and value of checking out assumptions; (g) bullying versus kindness; (h) cliques and value of compassion; and (i) cultural acceptance and tolerance. Each story is presented via videotape as well as through separate printed books. Original, illustrative artwork is integrated throughout the video, books, and other curriculum materials. Following presentation of the videotaped story, teachers lead a group discussion to engage students in first defining the story’s main lesson points and then connecting the story’s character traits and lessons to their own experience. To reinforce the story lessons and character traits, classroom and homework activities are utilized (selected based on allotted time and teacher choice). For the 3-5 curriculum, teachers may choose from the following set of activities: (a) story mapping to enhance story comprehension; (b) creative storytelling to extend beyond the story and expand on the meaning of the story; (c) Write-On to analyze the story lesson through written language exercises; (d) Explor-A-Story to relate story lessons to students’ personal experience in language arts and writing exercises; and (e) Word Herd worksheets to link story lessons with other academic areas. Similar to the lower elementary curriculum, the combined activities last approximately 60 to 90 minutes per story and can be presented on multiple school days.

The current study was designed to assess the effectiveness of the LifeStories for Kids (K-2 and 3-5) school-based curricula. This evaluation presents pre- and postintervention findings for teacher-reported behavioral functioning (direct and indirect aggression, peer victimization, immature-impulsive behavior, and prosocial behavior), by contrasting data from two schools implementing the intervention with two control schools. We expected both LifeStories for Kids programs to improve the school-based behavioral functioning of students participating in the intervention.

All K-5 teachers (n = 102) in four elementary schools taught the grade-appropriate LifeStories for Kids curriculum to their students (n = 2,205). To recruit schools for the study, information describing the project and notifying schools that free curricula for all classrooms would be provided for participation was sent to area elementary schools in one public school district in central North Carolina. The first four schools to express interest at all levels (administrative, teacher, counselor) were selected for participation. Of these four, schools were assigned randomly to either fall or spring implementation onset. Therefore, two schools began the curriculum in fall 2006 (intervention) while two schools were not to implement the curriculum until spring 2007 (wait-list control).

In August, parent information letters describing the curriculum and evaluation methods were sent home with every student in the four public elementary schools. While all students participated in the curriculum, parents were asked to indicate whether their child had permission to participate in the study. Permission forms were returned to classroom teachers. Of the total pool of 2,205 Kindergarten through fifth grade students, parental consent was obtained for 1,975 (90%) of students. The students receiving parental consent were evenly distributed by sex (50% girls, 50% boys). The racial distribution (as reported in archival school records) was 44.6% White, 28.9% Black, 14.1% Hispanic, 6.4% Asian, and 4.7% multiple groups. Demographic information by curricula and intervention group is presented in Table 1.

Fifty (2.5%) of the 1,975 students at Time 1 did not participate in the evaluation at Time 2 due to exiting the school or extended absence from school. The level of attrition was similar in both grades K-2 and 3-5 (2% for both). Prior to conducting analyses of LifeStories for Kids effectiveness, the participants without data at Time 2 were compared to those with complete data in each age group to examine possible attrition bias on sex, limited English proficiency (LEP), intervention group assignment, and initial levels of the outcome variables. Participants in grades K-2 with data missing due to attrition were rated as more peer victimized (p < .01), but did not differ on other variables. Participants in grades 3-5 with data missing due to attrition were rated as displaying more direct aggression (p < .05), peer victimization (p < .05), and immature-impulsive behavior (p < .01). In addition, they were rated as displaying less prosocial behavior (p < .01) and were more likely to be boys (p < .01), but did not differ on other variables.

As part of a comprehensive program evaluation, teachers in the four elementary schools were asked to rate the social and behavioral, functioning of their students. Teachers completed a modified version of the Teacher Checklist (Terry, Underwood, & Coie, 1994). Specifically, teachers were presented with a behavioral description and rated the degree to which that description was true of each child in his/her class. Ratings were made on a 7-point scale ranging from a low of 1 (never true of this child) to a high of 7 (almost always true of this child). Items drawn from the original teacher checklist were combined with additional items to broaden the behaviors sampled in each of the measured scales: direct aggression (physical and verbal), indirect aggression, peer victimization, immature-impulsive behavior, and prosocial behavior.

A confirmatory factor analyses of the revised measure yielded acceptable model fit [χ² (125) = 1040.19, p = .00; CFI = .94, RMSEA = .06; SRMR = .04], supporting the general factor structure by meeting most of the Hu and Bentler (1999) recommendations for cutoff criteria: CFI ≥ .95, RMSEA ≤ .06, and SRMR ≤ .08. In addition, factor loadings on all items were statistically significant (p = .00) and large in magnitude, supporting that each item is an adequate indicator of the associated scale construct. Further, items within each scale demonstrated adequate internal consistency. Scale items and detailed psychometric information is presented in Table 2. For each scale, an average score was calculated at each time point.

Information regarding student ethnicity, sex, and LEP status was obtained from the district database for purposes of sample description and covariate adjustment.

In August 2006, teachers in schools implementing LifeStories for Kids during the fall semester attended a 3-hour training workshop that gave an overview of the program, described the specific program stories and associated character traits, and provided opportunities to practice the presentation and classroom activities. During the 9-week intervention period, teachers implemented the program, devoting one week to each story lesson. Over the course of the week, teachers averaged between 60 and 90 minutes of total implementation time, spanning two or more days. The first day consisted of watching the story, completing a class discussion, and engaging in the creative storytelling activity. The second day consisted of completing at least one additional in-class activity (selected from the available options by the teacher) and one homework activity assignment to reinforce the story lesson. During the course of the intervention, project interns (graduate students in psychology-related fields) and project coordinators (senior research staff) were available to assist teachers with questions regarding program implementation and provide follow-up technical assistance as needed.

To encourage adherence to the intervention protocol, teachers filled out short intervention fidelity forms each week to document the specific intervention activities completed during the week along with any alterations (omission or additions) to the program material. In addition, implementation by each teacher was observed multiple times during the semester by program staff to ensure that the intervention was implemented as designed. As determined by teacher fidelity checklists and classroom observations, K-2 teachers implemented an average of 77% of intervention activities (SD = .15), whereas 3-5 teachers implemented 64% (SD = .21). As expected, the distribution of fidelity scores for both grade levels was negatively skewed (i.e., many teachers had high implementation scores and few had low). This degree of program fidelity, although less than desirable, is similar to the levels reported in studies of other effective classroom-based interventions (e.g., Botvin, Griffin, Diaz, & Ifill-Williams, 2001; Tappe, Galer-Unti, & Bailey, 1995; Tricker & Davis, 1988).

The same procedures were used for the pre- and postintervention evaluation. In October 2006 and January 2007, teachers were individually interviewed by a trained program staff member via a standardized interview script. Interviews lasted between 40 and 60 minutes on average. Teachers also completed a brief questionnaire following the interview. Teachers received a small gift (candy bar and $25 gift card) in appreciation for their participation.

To examine the impact of LifeStories for Kids program implementation, a quasi-experimental design (covariate adjustment for prior differences between groups) was utilized. Although the intervention was assigned at the level of the school, the analyses were structured at the student level. Due to the limited number of schools in the study, the nonindependence in the design (i.e., assignment of students, as clustered in schools, to the intervention) cannot be addressed in hierarchical modeling, potentially resulting in inflated estimates of statistical significance and effect size (Murray et al., 1994). Consequently, this study should be considered a preliminary study in support of a larger study with enough schools to provide adequate statistical power in hierarchical linear modeling to partition variance at the level of the student as well as the school to more appropriately assess the effects of intervention participation.

Analyses were conducted separately for the two curricula (K-2 and 3-5). Differences between students at postintervention were examined by regressing post-intervention variables on their corresponding values at baseline, with all outcome measures allowed to covary at baseline and posttest (Cohen, Cohen, West, & Aiken, 2003). Sex and LEP status were included as covariates to correct for the differences in composition between the groups at both pre- and postintervention and to improve statistical precision (Porter & Raudenbush, 1987). Intervention participation was incorporated in analyses as a dummy-coded variable (0 = control, 1 = intervention) using a multiple indicators and multiple causes model formulation to test for group mean differences (Jöreskog & Goldberger, 1975). Consequently, group differences were indicated by statistically significant parameter estimates for the regression of postintervention measures, with variance in the outcome variable explained by baseline values, sex, and LEP removed, on the dummy-coded intervention variable [an analogue to MANCOVA in structural equation modeling (Kano, 2001)]. To determine the multivariate significance of the intervention effect, the fit of a null model, with the paths from the intervention variable to the outcome variables constrained to equal zero, relative to a model with the paths freely estimated was compared (χ² difference test; Kline, 2005). Standardized effect sizes for statistically significant differences were obtained by dividing the unstandardized regression coefficients for the intervention contrast by the square root of the residual variance of the outcome measure, as recommended by Hancock (2001).

All statistical analyses were conducted in Mplus 4.2 (Muthén & Muthén, 1998-2007) with the MLR estimator (to make the chi-square test and standard errors robust to nonnormality; Yuan & Bentler, 2000), using Full-Information Maximum Likelihood (FIML) estimation to account for missing data. Based on the assumption that data missing due to attrition could be predicted by the individual’s scores on measured variables at prior measurement occasions (satisfying the assumption of data missing at random), FIML estimation is an appropriate technique to address missing values without the parameter and standard error biases and loss of statistical power that can be incurred through other techniques, such as listwise deletion or mean substitution (Allison, 2003; Raykov, 2005; Schafer & Graham, 2002). FIML is commonly used to address data missing due to attrition in longitudinal studies (e.g., Gottfried, Marcoulides, Gottfried, Oliver, & Guerin, 2007). Data are generally considered to be missing at random if the probability that the data are missing is predictable by observed variables (i.e., attrition can be related to observed responses on prior occasions); in contrast, data are considered to be missing completely at random if attrition is independent of observed responses at every measurement occasion (Schafer & Graham, 2002). Based on the observed initial differences in outcome measures by attrition status, the assumption that data are missing completely at random is not met in this sample, prohibiting the use of listwise deletion to address missing data (Allison, 2003; Schafer & Graham, 2002).

Evaluation of overall model fit was determined by investigation of the comparative fit index (CFI), the Standardized Root Mean Square Residual (SRMR), and the root mean square error of approximation (RMSEA). The reporting of multiple indices of fit is recommended, and these are among the most frequently reported to indicate overall model fit (Kline, 2005). Hu and Bentler (1999) recommended the following criteria for model acceptance on each fit index: CFI ≥ .95, SRMR ≤ .08, and RMSEA ≤ .06; however, considerable controversy exists regarding the use of global fit indices and specific criterion levels to evaluate model fit (see McDonald & Ho, 2002). Likelihood ratio chi-square values are also reported based on SEM convention, but they were not interpreted as primary indices of fit due to the sensitivity of the statistic to large sample size and the resulting potential for overrejection of adequate models (Kline, 2005).

The means and standard deviations of all outcome measures are presented in Table 3. In addition, all regression coefficients for the K-2 and 3-5 curriculum analyses are presented in Tables 4 and 5, respectively. Results are reported in terms of unstandardized regression coefficients due to the complication that arises when using nominal (i.e., dummy coded) variables. Specifically, standardized coefficients are difficult to interpret due to the variation in the magnitude of standardized regression coefficients based on the relative group size (Cohen et al., 2003). Readers are encouraged to refer to the magnitude of the standardized effect size (d) to determine the magnitude of the intervention effects.

The overall model for grades K-2 met most criteria for model acceptance [χ² (20) = 114.55, p = .00; CFI = .98, RMSEA = .07; SRMR = .06], permitting the interpretation of model parameter estimates. The full model fit the data significantly better than the null model, in which intervention effects were constrained to equal zero on all outcome measures [χ²_diff (5) = 44.30, P = .00], demonstrating a statistically significant multivariate intervention effect. Inspection of the model parameter estimates indicated improved direct aggression (B = -.14, P = .00) and prosocial behavior (B = .17, p = .01), while controlling for initial levels on these variables and the covariates, for students participating in the intervention. At pre-intervention, differences by intervention group, LEP, and sex were evident on the behavioral measures, justifying the inclusion of the covariates and the modeling of initial behavior differences by intervention status in the model. Boys had higher initial levels of peer victimization (B = .11, p = .03) and immature-impulsive behavior (B = .50, p = .00) and lower initial levels of prosocial behavior (B = -.32, p = .00). Students with LEP had lower initial prosocial behavior at Time 1 (B = -.26, p = .05). In addition, students participating in the intervention had lower initial direct aggression (B = -.19, p = .01), indirect aggression (B = -.33, p = .00), victimization (B = -.24, p = .00), and immature-impulsive behavior (B = -.19, p = .02). Differences by sex and LEP were also found on post-intervention behavioral measures, after controlling for initial values on these measures. Students with LEP had general reductions in direct aggression (B = -.11, p = .03) and immature-impulsive behavior (B = -.16, p = .04). Boys had general increases in direct aggression (B = .14, p = .00) and immature-impulsive behavior (B = .21, p = .00) and decreases in prosocial behavior (B = -.17, p = .01) during the study.

Controlling for the preintervention differences between the intervention groups and behavioral differences by sex and LEP at both pre- and postintervention provided a less biased estimate of the intervention effects. Specifically, the inclusion of the covariates improved statistical matching of the intervention and control groups and controlled for behavioral changes not related to intervention participation (i.e., behavioral changes related to sex or LEP status). Accordingly, students participating in the K-2 intervention had significant reductions in direct aggression (d = -.21) and increases in prosocial behavior (d = .17), as compared to students in the control group, with effect sizes of small magnitude (Cohen, 1988). No statistically significant differences were found by intervention status on indirect aggression, peer victimization, and immature-impulsive behavior.

The overall fit for the grades 3-5 model, although meeting most criteria for acceptance, had a questionable RMSEA value [χ² (20) = 205.12, p = .00; CFI = .97, RMSEA = .10; SRMR = .08]. The addition of a path from initial indirect aggression to post-intervention direct aggression significantly improved model fit [χ² (19) = 89.19, p = .00; CFI = .99, RMSEA = .06; SRMR = .06], and this model was used to test the intervention effects. Comparison of the full model with the null model (i.e., no intervention effects) supported a multivariate intervention effect [χ²_diff (5) = 20.19, p = .00]. Inspection of the model parameter estimates indicated improved direct aggression (B = -.11, p = .03) and immature-impulsive behavior (B = -.13, p = .01), while controlling for initial levels on these variables and the covariates, for students participating in the intervention. Similarly to the K-2 analyses, initial differences by intervention group, LEP, and sex were evident on the behavioral measures; thus, LEP and sex as covariates and initial behavior differences by intervention status were included in the model. Boys had higher initial levels ofpeer victimization (B = .19, p = .01) and immature-impulsive behavior (B = .70, p = .00) and lower initial levels of prosocial behavior (B = -.52, p = .00). Students with LEP had lower initial prosocial behavior at Time 1 (B = -.36, p = .02). Although students participating in the intervention had no statistically significant differences in initial behavioral functioning, a statistical trend toward higher initial indirect aggression for the intervention group was present (B = .15, p = .06). Differences by sex, but not by LEP status, were found on postintervention behavioral measures, after controlling for initial values on these measures. Boys had general increases in direct aggression (B = .21, p = .00) and immature-impulsive behavior (B = .15, p = .00) and decreases in prosocial behavior (B = -.19, p = .01) during the study.

Similarly to the K-2 analyses, the inclusion of covariates was supported to improve the matching of the intervention and control groups and to control for changes in behavioral functioning related to sex and LEP status. Students participating in the 3-5 intervention had significant, small reductions in direct aggression (d = -.15) and immature-impulsive behavior (d = -.19) as compared to students in the control group. No statistically significant differences were found by intervention status on indirect aggression, peer victimization, and prosocial behavior.

This preliminary, quasi-experimental evaluation was designed to document the effectiveness of the LifeStories for Kids program as a universal, preventive CE program for children in school settings. Based on the evaluation results, students participating in the LifeStories for Kids curriculum for grades K-2 experienced positive change in direct aggression and prosocial behavior in comparison to students not participating in the program. In addition, students participating in the LifeStories for Kids curriculum for grades 3-5 experienced positive change in direct aggression and immature-impulsive behavior in comparison to students not participating in the program. Consequently, both the K-2 and 3-5 curricula were found to be effective for positively impacting students’ social behavior within the school setting.

Although support for effectiveness was found, the study has limitations. First, only teacher data were examined. Teachers were not blind to condition, and teacher ratings of students could have been biased by awareness of their participation as intervention or control schools. Further explorations of effectiveness incorporating multiple information sources (parent, self, and peer report) in combination with school academic and discipline records are necessary to investigate the consistency of effects across informants. Second, due to the small number of schools in the study, statistical power did not permit direct modeling of the non-independence between the level of random assignment (schools) and the level of measurement (students), likely leading to some overestimation of effect size and statistical significance (Murray et al., 1994). Furthermore, the small number of schools was not adequate to minimize initial differences on behavioral measures and demographic variables by randomization between the intervention and control groups, necessitating the inclusion of covariates (i.e., sex, LEP status, initial behavioral differences between the intervention groups) to control for factors, other than intervention, that could affect the observed outcomes. Additional studies of intervention efficacy with adequate power to address nonindependence due to clustered randomization in hierachical modeling are encouraged (see Raudenbush & Byrk, 2002). Also, the pre- and postdesign does not address whether reported effects persist over time, which longitudinal designs could address.

In this study several strategies were employed to promote intervention fidelity (i.e., the use of a standardized manual, formal training, and ongoing support), and intervention fidelity was assessed by teacher-reported checklists and classroom observations. The degree of assessed intervention fidelity, although not atypical for school-based interventions, could result in underestimation of intervention effectiveness. In a meta-analysis of school-based interventions, Gresham, Gansle, Noell, Cohen, and Rosenblum (1993) reported a significant relation between intervention fidelity and effect size, and consideration of intervention fidelity is important to maximize effects both during initial research studies and as consumers implement effective programs in other settings (Domitrovich & Greenberg, 2000). In addition to teacher training and support for teachers during the intervention, teacher buy-in and investment in the intervention implementation in the school are also important to encourage fidelity. The factors leading to teacher buy-in, however, are somewhat at odds with researcher concerns with fidelity. Specifically, teacher discretion in covering specific lessons and control over the structure of classroom implementation have been identified as crucial for teacher buy-in for school based interventions, leading researchers to describe program flexibility and adaptation as necessary for successful implementation (Greenberg et al., 2001; Ringwalt et al., 2003; Turnbull, 2002). Consequently, the identification of the most important components of the LifeStories for Kids program and collaboration between the intervention developers, teachers, and school administrators to develop acceptable modifications and adaptations is encouraged to promote greater intervention fidelity, particularly as the program is adopted in new settings.

Of theoretical interest, changing actual student social behavior by implementing CE programs is difficult in comparison to increasing knowledge or modifying attitudes (Berkowitz & Bier, 2005). Specifically, extant CE programs are more effective at impacting the understanding and knowledge of character than in impacting student behavior (Berkowitz & Bier, 2005). The LifeStories for Kids program, however, demonstrated changes in several areas of student social behavior, strengthening its contribution to the range of available CE programs. The Character Education Partnership reports that aggressive behavior, school behavior, and prosocial behavior were positively impacted in only 48%, 45%, and 43% of reviewed studies of 33 CE programs, in contrast to 87% of studies for character knowledge (Berkowitz & Bier, 2005). More important, the results of this evaluation can provide the foundation for a larger study to address the call for additional studies demonstrating the effectiveness of CE programs in the school setting through experimental, control-group designs (Berkowitz & Bier, 2005; Leming, 1997).

In addition to supporting general program effectiveness, this evaluation further supports the effectiveness of storytelling as a viable method for character education. LifeStories for Kids was designed to incorporate storytelling to capture children’s attention and imagination to make complex social skills, character choices, and social dilemmas more concrete and recognizable for children (Brightwood & DeRosier, 2007). The stories presented in LifeStories for Kids provide examples through which the character traits can be explored and assist students in linking the stories to personal experience. Subsequent discussion and classroom activities, incorporating interactive storytelling and role-playing, provide opportunities to further practice the skills targeted in the stories and to continue the process of personal reflection on the stories and character traits. The method of storytelling employs the power of oral storytelling to pique student interest, to explicate the problem-solving process through character actions, and to internalize lessons through discussion and interactive activities.

In summary, the LifeStories for Kids program was found to be effective for improving prosocial skills and reducing problematic behavior in elementary school students. This quasi-experimental study supports the use of this program for helping children build skills for social situations and decrease negative social behaviors. Consequently, the LifeStories for Kids program is recommended as a universal CE program in elementary school settings.

This research was supported, in part, by a grant from the National Institute for Mental Health (5R44MH07071-03). We wish to thank the staff and students of the Wake County Public School System for their cooperation and support in the implementation of this research project. In addition, we wish to thank Dr. Janey McMillen, Natalie O’Brian, and Dr. Melanie Wilson for their invaluable efforts as project coordinators as well as the many interns of the 3-C Institute for Social Development for their assistance in successfully completing this research project.