Languages Across Borders: Social Network Development in an Adolescent Two-Way Dual-Language Program
by Amanda K. Kibler, Allison Atteberry, Christine N. Hardigree & April S. Salerno - 2015
Background/Context: Two-way dual-language programs have become an increasingly popular educational model in the United States for language minority and majority speakers, with a small but growing number of programs at the high school level. Little is known, however, about how adolescents’ social networks develop in the contexts of these programs.
Purpose/Objective: This study examines how a two-way, dual language enrichment program for Spanish-language learner (SLL) and English-language learner (ELL) adolescents influenced students’ social networks with peers of different cultural and linguistic backgrounds.
Setting: The program took place in a south-Atlantic state at a suburban/rural high school that has substantial within-school linguistic segregation.
Population/Participants: Program participants included 20 students: 10 English-dominant learners of Spanish, and 10 Spanish-dominant learners of English.
Intervention/Program: The two-way dual-language program was a voluntary extracurricular program in which adolescent Spanish-dominant ELLs and English-dominant SLLs participated in collaborative and student-led bilingual activities designed to foster the sharing of cross-linguistic expertise and cross-cultural knowledge over a seven-month period.
Research Design: In this mixed-methods study, student-level Likert-scale data is analyzed quantitatively and supported through analysis of qualitative interview responses and observational field notes. Quantitative results compare ELL and SLL participants’ demographic and baseline social characteristics, before-and-after social networks, the changing nature of reported relationships over time as a function of language status, and magnitude of growth in relationships’ strength before and after the program. Qualitative results examine the qualities and conditions of these relationships and the conditions under which they developed.
Findings/Results: Results suggest that despite participants’ demographic differences, ELL and SLL students in the dual-language program reported building new, strengthened, and mutually recognized relationships, particularly with students of different language backgrounds who worked together within long-term collaborative small groups.
Conclusions/Recommendations: When students are provided with a carefully designed instructional and ecological context that provides authentic purposes for using language and building peer relationships, this research suggests that both ELLs and SLLs may be able to build linguistically integrated social networks.
Like schools around the world, those in the United States are undergoing a dramatic shift: While educational systems and structures have remained fairly stable, students in classrooms look and sound very different than they did even a decade ago. Although immigration has been a consistent feature of the American landscape (Salomone, 2010), more recent changes in U.S. schools compositions have led to a significant body of research examining academic and linguistic outcomes of English language learners (ELLs), also referred to as emergent bilingual (García & Kleifgan, 2010) students. Theory and research have also been concerned with the rate at which the United States squanders its multilingual resources (Hornberger, 2002): The vast majority of immigrant families lose their home languages by the third generation (Rumbaut, 2009), and many school-based foreign language programs for English-speakers have disappointing results. In parallel, ongoing linguistic (Callahan, 2005; Gifford & Valdés, 2006) and ethnic (Reardon, Grewal, Kalogrides, & Greenberg, 2012; Ryabov & Van Hook, 2007) segregation within and among schools suggest limited opportunities for adolescents of different language backgrounds to interact with each other in school settings. Our study builds on and extends prior work suggesting the importance of adolescent peer social networks to explore the potential of two-way dual-language programs to positively influence these networks, particularly for language-minority youth.
This mixed-methods study examines outcomes related to Languages Across Borders (LAB), an extracurricular two-way dual-language program in which adolescent Spanish-dominant ELLs and English-dominant Spanish language learners (SLLs) participated in collaborative activities designed to foster the sharing of cross-linguistic expertise and cross-cultural knowledge. We address how this language education program for adolescents influenced students social networks with peers of different cultural and linguistic backgrounds. Specifically, we address the following research questions:
How do students reported social networks change from a quantitative perspective? We present visual representations of the social networks before and after the LAB program. We also conduct analyses that estimate changes in the number and strength of connections among LAB participants, and we look for evidence that those changes happened for both SLLs and ELLs, across language backgrounds, and in ways they could have been related to program activities.
How do students social networks change from a qualitative perspective? We describe the development of students peer relationships in LAB, the qualities and conditions of those relationships, and to what extent those relationships are visible in observational data.
When considering LAB as a two-way dual-language setting, other research questions of equal importance include the extent to which students developed insights into themselves, language, language learning processes, and other people. We have analyzed these outcomes, which we call ethnolinguistic insights, in a separate paper (Kibler, Salerno, & Hardigree, 2014) in which we draw upon interview and observational data to explore ways in which the program provided opportunities for adolescents to recognize and ratify peer ethnolinguistic identities; understand connections among languages, their speakers, and their cultural practices; and appreciate their own and others ethnolinguistic resources. The current study builds upon these insights in order to examine an important means by which students gained those ethnolinguistic insights: the development of linguistically diverse peer social networks within the program.
PEER SOCIAL NETWORKS AND LANGUAGE LEARNING
SOCIAL NETWORKS AND SCHOOL SEGREGATION
For purposes of this study, social networks are defined as a system of relationships between individuals which channels, and is constituted by, social interaction (Palfreyman, 2006, pp. 356357). In other words, relationships can be seen as the connections between individuals through which social capital flows. In the case of two-way dual-language programs, more linguistically integrated social networks are hypothesized to develop among students from both language backgrounds through opportunities for peer interactions. English-dominant students learning Spanish, for example, would benefit from social networks with Spanish-dominant students, in that they are able to learn from the expertise, experiences, and knowledge of those students already fluent in the language; such a situation would apply similarly to Spanish-dominant students learning English.
Perhaps more than in other disciplines, social networks have been a popular topic in education since the explosive Coleman Report (Coleman et al., 1966) first began to link individual academic attainment with average peer attainment. As explained in the American Educational Research Association (2012) Amicus Brief in support of the respondents in Fischer v. University of Austin et al., research has consistently documented educational benefits of student body diversity. American schools today, however, are increasingly resegregated by race, ethnicity, and language (Orfield, Losen, Wald, & Swanson, 2004), and microsegregation within otherwise unsegregated schools is also common, by race and ethnicity (Ryabov, 2011), previous academic achievement (Veronneau, Vitaro, Brendgen, Dishion, & Tremblay, 2010), and language status (Callahan, 2005). Scholarship, such as that presented in a recent Teachers College Record special issue Segregation, Desegregation, and Integration: From History, to Policy, to Practice (Diem & Brooks, 2013), attests to the historic and contemporary challenges of building and maintaining truly equitable integrated school cultures, and a growing body of research suggests that marginalized communities of students continue to be locked out of social access and advantage due to lack of access to peers with high social capital, a concept we discuss further below (Rizzuto, LeDoux, & Hatala, 2009; Ryabov, 2011; Ryabov & Van Hook, 2007). Such a pattern is particularly troubling given that integrated peer networks have been shown to bring all students a range of social and educational benefits, including promotion of norms for social integration, cooperation, academic success, and tolerance (e.g., Goza & Ryabov, 2009; Tezanos-Pinto, Bratt, & Brown, 2010).
For Latino students, there is reason to believe that building social networks with cross-linguistic and cross-cultural peers is particularly important, in that peer networks have been found to support positive academic outcomes for Latino adolescents (Woolley, Kol, & Bowen, 2009), and building networks through extracurricular involvement has been found important in encouraging Mexican-American youth to complete school (Ream & Rumberger, 2008). Insufficient connections can prevent Latino students from gaining entrée into school leadership clubs and classes, environments that build additional skills and connections which in turn support academic identities and experience (Gibson, Gándara, & Koyama, 2004). But when given opportunities to build cross-group peer networks, such relationships can have long-lasting, meaningful effects for Latino adolescents, in terms of their self- and cultural expression and academic opportunities (Lewis-Charp, Yu, & Friedlaender, 2004)
SOCIAL NETWORKS AND SECOND LANGUAGE (L2) LEARNING
Included among those students negatively affected by segregated social networks are students in the process of acquiring second (or additional) languages. Language-based segregation can be a feature of adolescent peer social networks, particularly if youth have low proficiency or limited experiences in the language(s) they are learning (e.g., Titzmann & Silbereisen, 2009). Wiklund (2002) applies such thinking in studying immigrant adolescents in Swedish schools. She found students built stronger social networks with native Swedish students in a school setting with fewer immigrants overall. Additionally, students who had social networks with more native Swedish-speakers or students from different immigrant groups than their own (with whom it might require use of Swedish to communicate) performed better on Swedish language tests. These findings suggest that environmental structures affect how adolescent language-learners build relationships and learn languages.
In the United States, Suárez-Orozco, Pimentel, and Martin (2009) found that students of various ethnic backgrounds who were newly arrived in the country benefited academically from expanded social networks, and similar patterns appear to hold for adolescent immigrants more generally: in a study conducted with 99 Spanish-speaking adolescent immigrants in New York City, a significant predictor of outcomes on English proficiency assessments included having three or more peers with whom adolescents either spoke English at least half the time or talked about topics that were academic in nature (Carhill-Poza, 2011). Such work suggests that immigrant students connectedness in peer networks (e.g., the number and language background of peers to whom they are connected) may impact both linguistic and academic outcomes. A concern for language-minority students, however, is also maintaining and developing home language proficiency in the face of pressures to assimilate into the dominant societal language, and just as stronger second-language networks support additional language learning, individuals from language-minority communities who have more and stronger social network ties within that community are also more likely to maintain that language (Milroy, 1987). Better understanding how to build strong networks in both languages is thus of great importance in supporting bilingualism, bi-literacy, and positive academic and social outcomes.
Two-way dual-language programs are conceptualized in this study as one means of promoting linguistically integrated social networks in ways that support beneficial outcomes for all participants. Research on such programs has documented positive linguistic, academic, and social outcomes for language-minority students (Rolstad, 2005) and Latino students in particular (Lindholm-Leary & Hernández, 2011), particularly in efforts to support bilingualism and bi-literacy. Simultaneously, for English-dominant students, these experiences do not mean separating students from high-achieving peers but instead allow for broadened social networks that incorporate diverse peers, including those with language expertise in language-majority students target language, and all the linguistic and academic benefits these diverse networks provide (Lindholm-Leary, 2012). In this way, connections offered by two-way dual-language programs are thought to be beneficial, though perhaps differentially so (Kibler, Salerno, & Hardigree, 2014; Bearse & de Jong, 2008; de Jong & Bearse, 2011), for diverse participants.
In order to capture elements of LAB students networks, we draw upon social capital theory and ecological language learning theory, an innovative combination that brings together social and linguistic concerns, particularly for language-minority populations. Both of these theories are useful in understanding contexts for learning for a student population that is increasingly culturally and linguistically diverse but that is increasingly segregated by race, ethnicity, and/or language both among and within schools (Callahan, 2005; Gifford & Valdés, 2006; Orfield & Lee, 2005; Reardon et al., 2012; Ryabov & Van Hook, 2007), trends that can have negative consequences for learners (Rumberger & Palardy, 2005). We are particularly interested in those schools in which diverse students have shared access to facilities, in that they attend the same schools, but have very different access to learning, resources, and peers (all important elements of social capital) because of language background.
Stanton-Salazar (2004) describes social capital as the connections to individuals and to networks that can provide access to resources and forms of support that facilitate the accomplishment of goals (p. 18). Social capital is particularly well suited to the study of language-minority students, in that it is founded on a recognition of power dynamics and societal inequality [that] sensitizes us to the kinds of ecological contexts in which many minority and low-status youth grow up (Stanton-Salazar, 2004, p. 33). In this sense, the ongoing segregation of language-minority students within and across schools raises questions regarding the types of connections made available to these students as well as the access to social capital those connections provide. For language-majority students who might have greater access in segregated environments to certain kinds of social capital related to schooling, they are nonetheless deprived of many linguistic, academic, and psycho-social benefits that can only be gained in diverse environments. Granovetter (1983) suggests that social capital can be derived from both strong and weak connections, or ties, and that each type provides unique resources to individuals. Although strong ties might be seen as necessarily better in building social capital, weak ties can be considered more important than strong ties when bridging segregated social groups in terms of increasing access to new ideas, opportunities, and perspectives (Burt, 2004).
Language ecology, which serves as a metaphor to represent the dynamic interaction between language users and the environment as between parts of a living organism (Kramsch, 2002, p. 3), underscores these interrelationships between context and language learning. In other words, language learning does not simply occur through cognitive processes internal to the learner; it is an interactional accomplishment made possible by, and therefore inextricable from, the learners interaction with his or her environment, including his or her peers. From an ecological perspective, context is centralas van Lier (2002) refers to it, the focal field of study (p. 144)and cannot be disregarded in language education research, so that the differences between different classroom environmentsfrom teacher and peer characteristics, language expertise, and interactional patterns to instructional activities and even the physical environmentare far from trivial. In this sense, ecological approaches to language acquisition suggest that cognition is inextricably interwoven with [learners] experiences in the physical and social world (Leather & van Dam, 2003, p. 13). Ecological contexts are where language learners encounter opportunities for or inhibitions of action (Leather & van Dam, 2003, p. 4), what van Lier (2004) calls affordances for language learning, which are themselves embedded in the sociopolitical settings in which language is used (Creese & Blackedge, 2010). Two-way dual-language programs, although quite varied in their design, implementation, and quality, are nonetheless predicated upon the notion that ecological settings explicitly valuing and developing both languages can provide affordances for language learners in both languages, even within contentious societal contexts that traditionally devalue minority languages.
Drawing upon these theories, programs like LAB can be seen as potentially providing students with a language ecology that attempts to facilitate two-way dual-language learning and an institutional support that encourages bridges (Lin, 2001, p. 67) between social and linguistic networks for youth, particularly in linguistically segregated environments. In this way, it is hoped that such programs can:
Assist the individual, the peer group, and/or the larger peer community in developing empowering ways to participate in their multiple and culturally disparate worldsnot only for the purpose of individual educational success, but also as a means for preparing to become an effective and determined agent of social change and social justice. (Stanton-Salazar, 2004, p. 34)
The LAB curriculum emphasized daily student-led, teacher-facilitated Spanish- and English-language activities. Activities purposely drew from students varied cultural backgrounds, required use of both Spanish and English, and aimed to help students explore cross-cultural communication while fostering development of mutually respectful relationships. Collaborative structures provided venues for developing relationships and the reciprocal sharing of peer expertise, in which students acted as peer mentors for their cross-linguistic partners while also receiving mentoring and assistance to complete collaborative tasks. Activities were not worksheet-style exercises but rather required students to rely upon and interact at length with partners. Interaction was facilitated by students shared knowledge of both languages and desire to learn their second language. In these ways, LAB provided an ecological context in which both languages were used and valued, and which supported positive interdependence, risk-taking, and the building of relationships with linguistically and culturally different peers. Peer linguistic expertise made available through such collaborative interactions, such as writing bilingual childrens books together, is understood as an ecological affordance (van Lier, 2004) that can support language learning and broader ethnolinguistic awareness (Kibler, Salerno, & Hardigree, 2014).
Students began the extracurricular seven-month program in rotating, cross-language pairs that each included one ELL and one SLL. Early activities encouraged oral communication, such as a What Would You Do If? activity; students took turns responding to cards with controversial scenarios like, You just started working at a fast-food chain. A co-worker is stealing from the register and blaming it on you. Students would answer the question in their second language, and their cross-language partner would respond and prompt with further questions. Other activities included inventing oral narratives, comparing familial dos and donts, and swapping trabalenguas, or tongue-twisters, among other activities. Later in the program, student pairs and groups engaged in more literacy-focused tasks, including a multiple-day activity about the pre-/adolescent Mexican group, Los Vazquez Sounds, in which students first read and watched videos in both languages about the group and then worked with partners to create podcasts designed for various purposes, with half of the pairs recording podcasts in each language. Students then listened to podcasts and attempted to determine the purpose of each podcast before engaging in a whole-class discussion of them.
Midway through the program, students, teachers, and researchers collaboratively created a dual-language service project, in which small book groups wrote bilingual books for first-grade Spanish-speakers at a local elementary school. The structure and grouping for this project was particularly important: we anticipated that a combination of sustained, small group interaction among students from both language backgrounds and carefully structured collaborative activities would be a primary mechanism of social network development. Student groups of four to six were designed to have roughly equal numbers of ELLs and SLLs in each group, and in order to build upon existing strengths developed in the first half of the program, we allowed students to select the top three ELLs and top three SLLs with whom they might like to work. We were able to honor some of their choices and create balanced ELL/SLL groups, but students were not given complete control in selecting their group members. As a result they were invariably placed with multiple students outside of their selected peer network. In this four-month project, book groups engaged in multiple related tasks, such as: creating bilingual podcasts asking first-graders about their reading interests, reading and discussing bilingual childrens books, developing story-maps for their book, and creating their books bilingual text and layout. (For additional details regarding LABs structure, curriculum, and activities, see Kibler, Salerno, & Hardigree, 2014).
PARTICIPANTS AND SETTING
The LAB program, which took place at Hamilton High,1 a suburban/rural high school in a small town in a south-Atlantic state, involved 20 students (10 ELLs/10 SLLs), three Spanish teachers, and two English/ESOL teachers, meeting two or three times weekly over seven months of one academic year. Students, in Grades 912, were enrolled in either: (1) Spanish 4 or Advanced Placement Spanish for SLLs,2 or (2) the English for Speakers of Other Languages (ESOL) program for ELLs, with students at proficiency levels from two (beginning) to five (bridging) on the WIDA Consortiums six-point English Language Proficiency Standards scale. The teachers shared information about LAB with all their students in these settings, and 20 chose to participate. All ELLs spoke Spanish as a dominant language; all SLLs spoke English as a dominant language, or in the case of four participants, spoke English as well as other languages at home (Gujarati, Jamaican Creole English, Spanish, and Kashmiri/Urdu) and were classified as fluent in English when they enrolled in the school district. LAB sessions occurred in two Spanish teachers classrooms during a 30-minute tutoring time before school and occasionally after school as well. Teachers facilitated sessions on a rotating basis, so at least one teacher from each language background attended each session. Researchers provided the curriculum and were present at each session to collect data and to assist as needed. Researchers presence and social positioning inevitably influenced both students and teachers in the setting (see Kibler, Salerno, & Hardigree, 2014 for further discussion), but the confluence of quantitative, interview, and observational findings presented below provide a measure of reassurance regarding the trustworthiness of results.
Because LAB aspired to foster social relationships, the study design included multiple data sources to capture changing relational dynamics among participants, including demographic survey results, interview transcripts, audio-recordings of student interactions, written documents, and daily fieldnotes written by multiple researchers. The primary data source for this mixed-methods analysis is transcripts from student interviews conducted at the end of the LAB program. In these interviews, students constructed visual social maps of their LAB peers before and after the program: Each student was given headshot photographs of all participants and was asked to organize the pictures as a map, with themselves in the center, depicting how close the student felt to other LAB members at the end of the program. This process was then repeated, but the student was asked to change his or her map to reflect what it looked like before the program. (The visual maps are not used directly in the current analysis; however, they provided researchers with a key visual prompt for interviewing students.) For both the before and after situations, students reported if they knew each of their 19 peers, and if so how close they felt to each person using a Likert scale, in which students described individual relationships with peers as ranging from 1 = not close to 5 = extremely close, with 0 = no relationship. This data, along with basic survey information, provides the basis for the quantitative analyses described below. Self-report data necessarily offers a partial view but nonetheless fits into many survey-based approaches to social network data collection. The fact that these maps were retrospective is also a limitation of this data collection approach, in that students might tend to remember ties that did not exist (Casciaro, 1998; Casciaro, Carley, & Krackhardt, 1999), but the substantial growth of social ties and notable patterns within this growth suggest that this data collection procedure nonetheless provided meaningful data, which are supported by both interviews and observations.
Data used in the qualitative analysis consisted of post-LAB interviews with students in which they were asked to make additional comments about the peers whom they scored on the Likert scale and to reflect upon their experience in the program overall. (For the latter, non-Likert-scale questions, see Appendix A for the qualitative interview protocol.) All interviews were conducted in students choice of language (English or Spanish), with interviewers fluent in that language. Fieldnotes were also used to locate observational data relevant to trends mentioned in interviews.
In this section we first describe our framework for data analysis. We then explain quantitative measures and methods used to conduct relevant regression and visual descriptive analyses before turning to a description of the qualitative analysis that deepens evidence provided by the quantitative work.
Framework for Analysis.
Research that seeks to establish a causal link between participation in a program like LAB and specific outcomes requires that (a) participants be selected for participation in a random or representative way, (b) some study participants receive treatment (here, LAB participation) while others do not, and (c) the mechanism by which participants are assigned to treatment or control conditions are controlled or accounted for by the researcher. Clearly, aspects of the current study do not meet these criteria. For instance, ELL and SLL students were recruited from a particular school based on availability and willingness. We are less concerned about whether the sample represents the school population, since the point of the current study is not to make generalizations to the entire school, but to demonstrate the feasibility and potential for programs like LAB to influence cross-linguistic and cross-cultural relations. Certainly it is possible that LAB volunteers were more likely predisposed to establish new relationships with others unlike them, so we cannot speak more broadly to how all ELL and SLL students would respond to a program like LAB. SLL participants frequent placement in college-bound tracks is also not representative of the entire SLL population at the school, though this was common among SLLs in advanced Spanish courses.
Although the current study does not include a traditional control group, we have valuable information about what social relationships would look like in the absence of LAB. These students generally attended the same schools for multiple years prior to LAB and had time to establish relationships with one another if they were going to arise organically. As we will see, many of the students did not know each other beforehand. Given the tracked nature of the school, we have little reason to suspect these students were likely to form new and deep relationships with one another in the absence of LAB. Still, it is possible that some other unobserved factor took place simultaneously with LAB and caused changes in relations between students that we might misattribute to LAB. However, we will also see that patterns of relationship development observed among these 20 students correspond to LABs specific intentions and structure.
Finally, it is important to address the issue of statistical power for our quantitative analysis. With only 20 students, it is difficult to estimate coefficients that are statistically distinguishable from zero. However, we use regression analysis to simply describe observed patterns in the population of participants, small though that population is. In some cases, we focus more on the meaning and magnitude of observed findings rather than strict statistical significance. In analyses at the dyad relationship-level, we have fewer concerns about sample size because there are many more potential relationships than individuals.
Measures for Quantitative Analysis
Student-level Likert-scale data underlies the quantitative analysis in the current paper. First, we constructed graphic visualizations of social networks both before and after LAB. Second, we constructed both student-level and relationship-level datasets for use in regression analyses to describe patterns in relational changes over time, as well as which kinds of relationships experienced most change.
In social network analysis, any two members of a network are called a dyad. Relationships between them are often called ties, and an existing tie is said to be realized. Some ties do not exist; that is, while Students 10 and 11 could know each other, they may report that they do not (tie = 0). In this dataset we not only know whether a tie exists, but we can also describe the strength (or value) of that tie using Likert-scale data reported by students (tie = 1 to 5 with 0 for no relationship reported, making six levels total). It is also important to note that ties have direction in that an individual reports ties out to others while others also report their ties in toward the individual. A student could report relatively few ties (or weak ties) out to others; however, others could view the student differently and report many ties in to the student. Thus data is considered non-symmetric.3
We can use valued, non-symmetric tie strength data to examine the overall network, changes for particular students, and changes for certain kinds of relationships. At the group level, network density is measured by dividing the total number of ties individuals report by the total number of possible ties. At any given time, each of the 20 students can have a possible tie with each of the other 19 students, giving a denominator of 380 possible ties. When many potential ties are realized, a network is characterized as dense, suggesting that its members interact a great deal.
At the tie or dyad-level, each tie has characteristics, and these characteristics serve as some of our primary explanatory variables of interest. For example, a tie can exist either within the same language group (a.k.a., a language match: ELL-to-ELL or SLL-to-SLL), or across language groups (a language non-match: ELL-to-SLL). Each tie also occurs at a certain time, either before or after LAB, so a given tie between Students 11 and 12, for example, might change over time.
Methods for Quantitative Analysis
In this section, we describe how we use quantitative measures to answer the following specific sub-questions.
(1a) How do ELL and SLL participants compare on basic demographic characteristics and baseline social characteristics? It is important to acknowledge that ELLs and SLLs differ in important ways aside from their dominant language. We present mean descriptive statistics separately for ELLs and SLLs and focus particularly on differences of demographics, language, and base state of social relations reported by students.
(1b) What do social network maps look like before and after LAB? Using UCINET (Borgatti, Everett, & Freeman, 2002) software, we created visual diagrams of LAB social networks before and after the program. These graphs illustrate individuals as geometric shapes and the ties among them using arrows of different thickness to indicate relationships reported strength. Although the graphs two-dimensional rendering distorts to some extent geodesic distances among individuals, they allow for a visual examination of networks relative density over time.
Table 1. Pre-LAB (Baseline) Demographic and Social Descriptives, by Language Program Status
(1c) How do changes in number and strength of connections compare for ELLs vs. SLLs during the program? LAB is intended to increase and strengthen relationships between students of different language backgrounds. To examine whether ELLs and SLLs mean out-tie scores were significantly different from each other before and after LAB, we ran the following difference-in-differences4 regression model on the forty observations of twenty students’ mean reported out-ties both before and after LAB:
Out_meantiesit = β0 + β1 (elli) + β2 (postit) + β3 (elli × postit) (1)
In Equation 1, we regress our outcome of interest, student is mean reported out-ties at time t, on a set of indicator variables including an "elli" indicator variable coded as 0 for SLL students and 1 for ELL students, a "postit" indicator variable indicating whether a given observation occurred pre- (0) or post- (1) LAB, and an interaction of these two. The intercept, β0, represents the mean out-tie strength for SLL students before LAB began. The coefficient on the ELL indicator, β1, indicates the mean difference in outcomes between ELL and SLL students before LAB. From this, we can learn more about preexisting social connections for ELL and SLL students. The coefficient on the "post" indicator variable, β2, captures the predicted positive difference in the mean strength of out-ties for SLL students at the end of LAB. We hypothesize that students should exhibit growth in reported tie strength (a positive coefficient). We are particularly interested in whether ELL students appear to have similar increases in mean tie-strength during LAB to SLL students, or whether these two groups have different experiences; β3 captures this difference in pre/post growth between ELLs and SLLs. What is of most importance to this research question is the difference in pre-/post-growth for within- and between-group relationships. If LAB bridged ELL and SLL networks, we expect to find a positive coefficient on the interaction between language-match and the pre/post indicator, which would suggest that cross-group relationships were particularly fostered by LAB participation.
(1d) Is there more growth in relationship strength for student pairs of similar or different language backgrounds? We employ a similar regression approach here to that used in 1c; however, we reorganize the data into a dyad-time level structure. In this new dataset, each of the twenty students has a possible tie with each of the other 19 students at two different times (pre-/post-LAB), giving a denominator of 760 possible ties. Each potential dyad from student i to j has a reported out-tie value: students either report no tie (outcome = 0), or they report a strength of the relationship between 1 and 5. This reported out-tie becomes the outcome of interest in Equation 2:
OutTieijt = β0 + β1 (BetweenELLij) + β2 (postijt) + β3 (BetweenELLij × postijt) (2)
Here, "OutTieijt" represents the reported tie from student i to student j at time t. Similar to Equation 1, we then model this outcome as a function of two indicator variables and their interaction. The new variable here is "BetweenELLijt" an indicator variable equal to 0 if student i and j have the same dominant language (either both are SLL or both are ELL), and equal to 1 if student i and j have different dominant languages. This variable indicates whether a given reported tie exists for members within the same language group, or between groups. As before, "postijt" indicates whether a given observation is related to pre- (0) or post- (1) LAB. In Equation 2, β0 is the mean out-tie for within-group relationships in the pre-period. This captures how well students of the same language designation knew one another before LAB began. The coefficient on the “between” indicator, β1, tells us whether there was a significant difference in mean out-ties when comparing within- and between-group relationships in the pre-period. We hypothesize that we will observe a negative estimate for β1 since the program presupposed that, in the absence of LAB, students of different language backgrounds are less likely to form positive school-based relationships. β2 captures the growth in mean out-ties from pre- to post-LAB for within-group relationships. One expects that the LAB program will bring all studentseven those with similar backgrounds and who already knew one anothercloser. What is of more importance to this research question is our estimate of β3, which represents the difference in pre-/post- growth for within- and between group relationships. If the LAB program is working as intended, we expect to find a positive coefficient here, which indicates that cross-group relationships were particularly fostered by LAB participation.
(1e) Did students within the same project groups form stronger relationships during LAB? One concern about the analysis of the LAB social networks is that it is difficult to prove that the LAB experience itself was actually responsible for increased tie-formation. Instead, simple exposure may cause increased familiarity among students. In other words, one might argue that no particular structured activities are necessary to form cross-language relationshipsthat one could just put 20 previously unacquainted students in a room and they would become friendly over the course of the school year. If all tie formation occurred as a result of sheer proximity, then we would not expect any systematic patterns in terms of greater tie strength growth during the year.
If, on the other hand, tie-formation was related to the core activities of LAB, then we would expect that students within the same bilingual book project group would exhibit greater growth than students in different groups.5 The service projectin which SLLs and ELLs were mixed into five small book groups of four to six students eachwas one of the primary activities in the LAB program. The creation of these books was central to the objectives of the LAB program, because it was explicitly bilingual in both process and product, required expertise in oral and written forms of both languages, involved sustained interaction in a group of linguistically diverse peers, and provided a meaningful context for interaction.
To explore this, we employ a similar regression approach here to that used in 1d using the dyad-time level structure (again, for a denominator of 760 possible ties). Each potential dyad from student i to j is either assigned to work within the same project group, or in different groups. The new predictor of interest, then, is an indicator variable, "sameprojectij" which equals 1 if student i and j were assigned to work within the same project group, and 0 if the two students were in different project groups. In Equation 3, we model whether tie formation is organized as a function of project group, and whether there is more within- or between-project group improvement over time.
OutTieijt = β0 + β1 * (sameprojectij) + β1 * (postijt) + β1 * (sameprojectij x postijt) (3)
(1f) How mutual are observed student relationships pre- and post-LAB for ELLs and SLLs? This research question probes beyond formation of new ties to examine strength and quality of the ties. We posit that strong relationships are characterized by mutual and common understanding of relationships strength. When two individuals view their relationship differently, it suggests either a relationship imbalance or a miscommunication between individuals. As a relationship develops over time, individuals come to a more shared understanding of their closeness. Here we use the same basic regression framework to examine mutuality of relationships as our outcome of interest, and we model it again as a function of time, and language-match status, and an interaction of the two:
Mutualijt = β0 + β1 (BetweenELLij) + β2 (postijt) + β3 (BetweenELLij × postijt) (4)
In Equation 4, the outcome is an indicator variable which is equal to 0 if the relationship between student i and j at time t is “mutual”: That is, "Mutualijt" was equal to 1 if the reported tie strength from student i to student j matched the reported tie strength from student j to student i. For example, if student i reported a tie strength of 3, then the relationship was designated as mutual if and only if student j reported a tie strength of 3 to student i. In this case, we are interested in whether same-language relationships were more likely to be mutual post-LAB (β2), but even more importantly whether relationships between students with different dominant languages became more mutual post-LAB (β3).
When examining mutuality, one important sample restriction must be made. Likely the best understood and mutually agreed-upon relationship tie is one that does not exist. In the current project, ELLs and SLLs were more likely to not know one another; in fact, all 34 relationships in which both parties agreed they did not know one another were between SLLs and ELLs before LAB. Thus, it appears that the most mutual relationships are those between SLLs and ELLs before LAB began. Because this is not the kind of mutuality of interest for this question, we dropped from this analysis the 34 dyads in which both students agreed they did not know one another. The regression results therefore speak to the mutuality of relationships pre- and post-LAB for ELLs and SLLs who reported knowing one another at the programs onset.
(1g) How much do relationships grow from pre- to post-LAB? It is important to examine not only whether relationships grew stronger on average, but also the heterogeneity in growth depending on the initial strength of the relationships. We simply tabulate each relationships growth by the initial strength of each relationship to demonstrate the transition matrix of relationships pre- to post-LAB. We then tabulate the post-LAB reported ties for those initially nonexistent relationships to see the distribution of relationships after the program. We do this for each of the six initial starting levels (0 through 5).
Methods for Qualitative Analysis
Qualitative analysis employed an iterative process of inductive and deductive coding (Coffey & Atkinson, 1996; Marshall & Rossman, 2011) through examination of students interview comments regarding relationships. Interview transcripts were read in their entirety by three of the co-authors, who each suggested inductive coding schemes using an open-coding (Emerson, Fretz, & Shaw, 2011) process. The first author compiled and synthesized suggested codes and aligned them with trends in existing research; codes were subsequently revised in collaboration with the other authors based upon corpus re-readings for disconfirming evidence. Inter-coder reliability was 83%; instances of coding doubt were resolved by consensus. After coding of interview transcripts, researchers used those same codes to search for both confirming or disconfirming evidence in fieldnotes, After recursively coding, re-reading, and refining codes, we organized students qualitative descriptions of their social relationships into the following three categories: new relationships, strengthened acquaintanceships, and out-of-class relationships. Appendix B includes full code descriptions, examples of interview data for each code, a sample coded interview excerpt, and a table that shows how observational fieldnotes and interview data were integrated for this analysis.
Results present evidence of adolescent ELLs and SLLs reported social networks with peers of different cultural and linguistic backgrounds before and after participation in a two-way dual-language education program. We turn first to the quantitative evidence, which is presented according to the six sub-questions listed above; we then explore the qualitative findings.
In what follows, we begin with a quantitative comparison of ELLs and SLLs before LAB (1a). We then present social network diagrams for students pre- and post- LAB (1b). Research questions 1c, 1d, and 1e examine the changing nature of reported relationships over time as a function of language status. Finally, we examine the magnitude of growth in relationship strength for all students from before to after LAB (1f).
(1a) How do ELL and SLL participants compare on basic demographic characteristics and baseline social characteristics? Table 1 results provide background for an analysis of ELL and SLL students demographic characteristics. It also compares baseline pre-LAB relationship descriptives, which are similar, though not the same across language groups. ELL students report slightly stronger mean ties than SLLs do before LAB (2.0 to 1.6, respectively, on a scale from 0 to 5). There is also some evidence of within-group homophily wherein students report more connection to students with similar attributes: SLLs report more ties to other SLLs, whereas ELLs report more ties to other ELLs. Differences also exist between groups in weighted GPAs (2.72 for ELLs and 4.35 for SLLs), and additional analysis of transcripts (see Kibler, Salerno, & Hardigree, 2014) suggests these students rarely enrolled in the same track of coursework, with SLLs in advanced tracks and ELLs in basic or remedial tracks.
(1b) What do social network maps look like before and after LAB? Figure 1 depicts social network diagrams for the 20 students before and after LAB. Here, visual cues such as shape, color, position, distance, and line thickness illustrate networks multiple dimensions concisely in one graphic. Two main takeaways are readily apparent. First, ELL/SLL segregation is immediately noticeable in the pre-LAB diagram, with thinner (weaker tie) and fewer lines among the groups; however, there is much more mixing between ELLs and SLLs in the post-LAB diagram, with thicker and more lines among groups, suggesting increases in both numbers and strength of ties from the first to second diagrams. One can quantify the degree of a networks social cohesion using a density measure, which for a valued network like these is the total of all values divided by the number of possible ties. Pre-LAB density is 1.79, while post-LAB density is 2.54, suggesting a meaningful increase in network density over time.
(1c) How do changes in number and strength of connections compare for ELLs vs. SLLs during the program? Because the LAB program met often and involved substantial interaction, all participants reported that they at least knew all of the 19 other students at the studys end (i.e., all possible ties existed). Due to this, we focus on modeling the variation in the mean strength, rather than number, of out-ties for question (1c). The results for this question are presented in Table 2, along with results for the following three research questions that use a similar regression framework. For all analyses using the regression method presented in Table 2, we can sum across various reported coefficients to obtain estimates of the mean outcome pre- and post-LAB for both SLLs and ELLs. All results reported are either directly from Table 2 or are derived by adding up the appropriate coefficients from Table 2 to represent the relevant group.
The intercept, β0 = 1.576 (p < 0.001) represents mean out-tie strength for SLLs before LAB. Recall that the outcome is based on a Likert scale in which students reported no tie (0) or an existing tie of strength 1 (least close) to 5 (closest). The coefficient on the ELL indicator, β1, was estimated to be 0.434 scale units (p > 0.05). This indicates that, on average, ELL students reported slightly stronger ties before LAB than SLLs; however, that difference is not statistically significant. The coefficient on the “Post” indicator variable, β2 captures the predicted positive difference in the mean strength of out-ties for SLLs at the end of LAB. This coefficient, β2 = 0.800 (p < 0.01), reports a significant increase in mean tie strength for SLL students during LAB: almost an entire scale-point increase. We are particularly interested in whether ELLs appear to have comparable increases in mean tie-strength during LAB, or whether these two groups have different experiences. β3 corresponds to this research question by capturing the difference in pre-/post-growth between ELLs and SLLs. The coefficient of -0.111 units is small and not significant. Taken together, we can interpret this as evidence that both groups had similar mean out-ties before the program and both groups experienced significant growth of about the same magnitude during LAB.
(1d) Is there more growth in relationships for student pairs of similar or different language backgrounds? Relevant results are shown in Table 2, Column 2. The intercept, β0, is the mean out-tie for within-group relationships in the pre-period. Before LAB, the mean out-tie for a pair of students with the same dominant language is 2.872 (p < 0.001). The coefficient on "BetweenELL," β1, represents the difference in mean out-ties when comparing within- and between-group relationships in the pre-LAB period. Before LAB, out-ties between two students with different dominant languages was 2.042 units lower than that of student pairs with the same dominant language. Therefore, the mean out-tie reported for between-group ties was about 0.83, a very low mean tie strength for cross-language relationships before LAB. β2 captures growth in mean out-ties from pre- to post-LAB for within-group relationships. For dyads of the same dominant language, ties were, on average, 0.394 units higher post-LAB. Again, we are most interested in the comparison of pre-/post-LAB growth for same-language versus cross-linguistic relationships, captured by β3. We find that dyads of different dominant languages grew even more than the same-language dyad relationships over the course of the program. The differential growth rate was 0.666 units (p < 0.001). This indicates that, while dyads of the same language increased from a mean reported out-tie of 2.87 to 3.27 (a 0.394-point increase), dyads of different languages experienced a 1.06-point increase from pre- to post-LAB, nearly triple the rate for same-language relations. It is interesting to note that, despite this strengthening of cross-linguistic ties during LAB, mean between-language tie strength was about 1.89 scale units after LAB. Therefore, on average relationships between students of different languages were still weaker than relationships of students with the same dominant language. Data are nonetheless consistent with the notion that LAB differentially targeted and fostered cross-linguistic relationships. Figure 2 visually depicts this finding. Again, we see that almost all relationships remained the same or grew stronger during LAB (flat or positive slopes). As expected, same-language relationships are stronger overall, and cross-linguistic relationships are especially weak in the pre-LAB period. The main finding is the steep slopes of the cross-group relationships for both ELL and SLL students; this is where most growth in out-tie strength was concentrated.
Table 2. Regression Results for Quantitative Research Questions 1c-1f
(1e) Did students working in the same project groups form stronger relationships during LAB? Results for question (1e) are shown in Table 2, Column 3. The intercept, β0, is the mean out-tie for between-group relationships in the pre-period. Before LAB, the mean out-tie for a pair of students in different project groups is 1.80 (p < 0.001). The coefficient on "SameProject," β1, represents the difference in mean out-ties when comparing within- and between-group relationships in the pre-period. Before LAB, out-ties between two students in the same project group were 0.028 units lower than student pairs in different project groups. However, this difference is small and not significant. This suggests that the project groups were likely not comprised primarily of students who already knew each other well. β2 captures the growth in mean out-ties from pre- to post-LAB for between-group relationships. Here, we find that the dyads that were in different project groups reported, on average, 0.613 units higher/stronger relationships post-LAB (p < 0.001). This suggests that students working in different groups do appear to become more familiar with one another from before to after LAB. However, of most interest is β3, which tells us the difference in pre/post growth attributable to being in the same project group. We find that dyads that were in same project groups grew more than the dyad relationships in different groups. The differential growth rate was 0.806 units.
Figure 1. Social Network Diagrams Before and After ALB Program
Figure 2. Pre/post Changes in Reported Mean Tie Strength (Out to Others) Within-group vs. Between-group
Therefore, while we see growth both within- and between- project groups, we see more than twice as much growth in reported out-ties for students who participated in the project together. While dyads that were in different project groups went from a mean of 1.80 to about 2.41 post-LAB (a 0.613 unit increase), dyads that were in the same project group increased from a mean 1.77 to about 3.19 post-LAB (a 1.42 unit increase). Although we acknowledge that this trend may be influenced to some degree by students partial input into the group-formation process, these differences in growth nonetheless suggest that the bilingual book project appears to be important to the formation of stronger cross-language ties throughout the school year.
(1f) How mutual are observed student relationships pre- and post-LAB for ELLs and SLLs? Analyses for this question are presented in Table 2, Column 4. Recall that the 34 relationships in which both parties agreed they did not know one another were dropped prior to analysis, so the sample size has decreased from the 760 potential relationships to 726 relationships for whom at least one of the two students reports knowing the other.
The intercept, β0, represents the expected proportion of same-language dyad relations that were mutual pre-LAB. In other words, about 30% of ties between students of the same dominant language were mutual (p < 0.001) pre-LAB. The coefficient on "BetweenELL," β1, represents the difference in proportion of ties that were mutual when comparing within- and across-language relationships in the pre-LAB period. Before LAB, relationships between students with different dominant languages were slightly less likely to be described as mutual; however, that difference was not significant (β1 = -0.095, p > 0.05). The coefficient on the “Post” indicator captures growth in proportion of ties that are mutual from pre- to post-LAB for same-language relationships. About 15.6% more same-dominant-language ties were mutually agreed upon post-LAB (p<.05). The interaction coefficient, β3, indicates that growth in mutuality was even larger for dyads who spoke different dominant languages: these dyads became even more likely to be mutual than the same-language dyad relationships over the program. While the percentage of dyads of the same-language that were mutual increased from 30 to 45.6 (a 15.6-point increase), the percent of different-language dyads that were mutual went from 20.5 to 46.1 (a 25.6-point increase) during LAB.
(1g) How much do relationships grow from pre- to post-LAB? To this point we have reported mean out-tie strength before and after LAB for ELLs, SLLs, students who have the same dominant language, those with different dominant languages, and the mutuality of those relationships. We conclude the quantitative analysis by simply examining how each relationship changed in strength from pre- to post-LAB by language group. Table 3 provides a sense of the extent to which individual relationships were deepened; cells highlighted represent ties that deepened over time.
For example, of the 83 cross-linguistic dyads (left panel) who initially reported no tie, 100% were reported as stronger post-LAB, and about 32.5% reported an increase of two or more Likert-scale points. Of those cross-linguistic ties that deepened, the most prominent shift was that of moving from a value of 0 (not knowing one another at all) to a value of 1 (a weak acquaintance). Yet, of the 84 cross-linguistic dyads (left panel) who initially reported a very weak (1 = least close) tie, about 48.8% reported an increase of two or more Likert-scale points. It is interesting to note that only 11 of 200 possible cross-linguistic relationships reported a tie strength greater than 2 before LAB. In contrast, 103 of 180 possible same-language relationships (about 57.2%) reported a tie strength greater than 2 before LAB. By comparing the row and column totals for language-matching versus language-non-matching relationships, we can see that relationship formation patterns differ for the two groups. Overall, reported relationships that did not previously exist formed during LAB, and existing relationships deepened particularly for cross-linguistic dyads.
Students growing relationships within and across language groups can be further illuminated by considering qualitative data. Three themes emerged from analysis of students comments about their changing relationships in LAB and the qualities and conditions of those relationships: Students connected with those they might not otherwise have met; they strengthened relationships with those who were previously acquaintances; and they described relationships as sometimes extending beyond LAB into other parts of school space.
Table 3. Transition Matric of Pre-LAB Reported Out-Tie Strength and Post-LAB Out-Tie Strength, by Whether Pair of Students' Dominant Language Matches
Some students said their LAB relationships were completely new, involving students they might have never known had it not been for LAB. This sentiment was prevalent in students explanations, regardless of language background. Yesenia, an 11th-grade ELL, explained, Pues, [las] experiencias fueron buenas, porque pude conocer a otras personas que no había conocido antes. / Well, the experiences were good, because I could meet other people I hadnt met before.6 Yesenias new relationships were also documented in fieldnotes: At the beginning of the program, she interacted almost entirely with other ELLs, but by the end she was observed greeting and chatting socially with not only ELLs and SLLs in her book-project group but also students in other groups.
Students interview comments suggest that such opportunities to build networks across language groups are rare in other school-based settings, particularly for SLLs, who tended to be in advanced-level courses with few ELL students. Emily, an 11th-grade SLL, described opportunities presented in LAB as a chance to integrate:
What have you liked about the LAB program?
Definitely the fact that we get to integrate [with] other students from [Hamilton] that I might not have in classes, but now I obviously know that they attend [Hamilton], and then thats also the just like broadening like the people you know here.
From Emilys perspective, classes are a primary means of building peer relationships, and LAB gave her opportunities to meet students with whom she did not share classes, including SLLs in different Spanish levels, students at different grade levels, or students in different academic tracks. Madison, an 11th-grade SLL, adds another layer to this issue. When asked what she learned about language in LAB, she explained: I think it was just neat working with somebody that doesnt, their first language isnt English because pretty much the classes that Im in, everybody normally has their first language which is English, so it was a different experience, which is neat. Here, Madison suggests a link between language designation and course-taking patterns, in that ELLs are rarely present in her (almost exclusively honors) classes. Despite these patterns, both Emily and Madison actively engaged with cross-language peers during LAB: Their book-project group, which included the two SLL girls along with two ELL boys (Alex and Javier), was one of the most lively in the program, as they often veered back and forth between LAB tasks and more social conversations.
Like Emily, Josh, a 12th-grade SLL, noted specifically that LAB contrasted with what he saw as the typical, more homogenous school context:
Well, what Ive liked is that I was able to meet new people, which I sometimes dont do, because its a pretty constant environment at this school. Like hanging out with the same people for the whole time, so it was just neat to be able to meet different people.
Josh, a quiet student, noted that LAB was a change in the constant [peer] environment, exposing him to new people and connections outside of his existing peer network. Although not the most outgoing student in the program, Josh contributed thoughtfully in paired and group activities, particularly in creation of the bilingual book. While Joshs statement appears to include both academic interaction as well as social connections (hanging out), Jacob, an 11th-grade SLL, was perhaps most explicit in describing social separation between his own peer group and Spanish-speakers at the school, which LAB appeared to bridge. When asked what he liked about LAB, he responded:
I enjoy interacting with others, but usually I wouldnt be able to.
And who would that be?
Well pretty much the Spanish-speakers because I usually dont, um associate myself with them because its just not, its not that I dont like that, its just that its hard to when they have their own group, and Im in mine. But I enjoyed meeting them.
Although Jacobs description of meeting them appears to suggest somewhat superficial relationships (or what Granovetter, 1983, calls weak ties), his comment nonetheless speaks to powerful social barriers between linguistically different peer groups and the possibilities of bridging those divides in LAB. Particularly notable in this regard was an extended social conversation captured in fieldnotes at the end of the year, in which Jacob and Isabella, a ninth-grade ELL in another project group, talked about the progress of their respective books.
Several students said LAB gave them opportunities to strengthen relationships with others they had known only superficially before. For example, Farida, an 11th-grade SLL, described this situation in her developing relationship with Roberto, an 11th-grade ELL in her book group, whom she moved from a one to a four on the quantitative measure of her social map. When asked what she learned about other people through the LAB experience, she explained:
I think I realized that . . . people are more than they seem? Like when you actually get to know them.
Yeah, tell me about that.
Just the same thing [I said before about] Roberto. Just cause I would see him in Spanish class, and it was just kind of, he would be doing his own thing, but when we got him in a [book] group and we started talking, then it was exciting, it was fun.
In this sense, Faridas relationship with Roberto grew from one of recognition to one of interaction, or from weaker to stronger ties, as Farida and Roberto engaged jointly in the book project. Farida further discussed the many jokes she shared with Roberto by the end of LAB, including a recurring playful joke by Roberto, also captured frequently in our fieldnotes, about Faridas crush on a prominent CNN broadcaster. (See Appendix B for additional supporting data.)
Students also described how growing relationships sometimes extended beyond LAB into other parts of their school lives. One way in which this extension occurred was through talking in the hallway. Throughout the data, students described how their relationships strengthened through reference to times they talked with new LAB friends in social spaces around the school. For example, Yesenia, an 11th-grade ELL, described her interactions with Ethan and Leah, two 12th-grade SLLs:
Like, maybe cuando ya vamos abajo, estábamos en los hallways o cuando nos mirábamos así como que ya nos íbamos o veníamos otra vez a la escuela, nos poníamos a platicar o cosas así. / Like maybe when we go downstairs, we were in the hallways, or when we see one another as well as [when] we were coming and going to school again, we would start talking or things like that.
Yesenia mentioned these interactions after describing how an original challenge in LAB for her was to interact with people she did not previously know. After LAB, she explained that these feelings changed:
Sí cambian. Porque ya cuando empiezas a conocer a las personas, tú dices: Así entonces sí le hubiera hablado antes- hubiera perdido una gran amistad si no le hubiera hablado. O cosas así. / They do change. Because when you start to get to know people, you say, So then, if only I had talked to him/her before- I would have lost a great friendship if I hadnt talked to him/her. Or things like that.
In this way, Yesenia indicated that her cross-linguistic relationships were becoming great friendships. This is typical of how students appeared to use talking in the hallway as a sign of relationships with others. In this case, Leah reciprocated this sense of relational development and discussed how talking in the hallway builds that connection. In discussing the new relationships she made, she explained, Just like you know getting to know people and being able to encourage them and say hey in the hallway. Its like, youre not like strangers anymore. While it is unclear how deep these relationships extended, talking in the hallway appears to be an important step in extending social networks and possibly helping students feel they belong at school.
Likewise, students discussed talking in other common spaces, such as at lunch or after school. Javier, a 10th-grade ELL, explained that he was surprised when Emily (an SLL) showed an interest in talking to him outside of LAB. When asked about his strongest memories of working with his book-group members, Javier explained:
Actually I used to see her after school.
She used to talk to me after school, and I talked to her. I thought she was
just gonna talk to me just when the LAB [happened]. But it wasnt just like that.
In this excerpt, Javier revealed a prior expectation he had about Emily and how her talking to him outside LAB altered that. Observational data show that after school Emily could generally be found practicing dance moves with cast members for the school musical or participating in other extracurricular activities; talking to Javier after school for her might have involved acknowledging him in front of other students linked to similar, popular after-school events. This recognition of each other in public school spaces, where students not part of LAB see this cross-linguistic acknowledgement, can be an important step in establishing peer relationships and challenging segregated norms.
Both quantitative and qualitative data describe mutually reaffirming patterns in SLL and ELL adolescents social networks. Although ELL/SLL segregation was noticeable in initial social network mapping, both number and strength of ties increased by programs end. Next, we find that both ELL and SLL students experienced out-tie growth of similar magnitude, suggesting extended and enriched relationships for both groups. Speaking to the unique affordances of two-way programs in highly segregated environments (such as the status quo at Hamilton High School, in which many English- and Spanish-dominant students simply did not know each other), cross-linguistic dyads grew at nearly triple the rate as same-language relations, although the former remained relatively weaker than the latter. Mutuality of relationships is likewise positive: cross-linguistic dyads were even more likely to be mutual than same-language dyads over time. And while we see growth both within and between book project groups, we see more than twice as much growth in reported out-ties for students who participated in the project together. Finally, this data suggest that new relationships were formed and existing relationships deepened, particularly in cross-linguistic dyads.
Qualitative data support these trends, while adding useful context for development of these networks. When talking about new relationships, students generally tended to talk more about development of cross-language rather than same-language relationships, in keeping with the notion of differential growth in relations across language groups and the ecological affordances of a two-way dual-language context. Students comments also suggest the realities of school segregation for adolescents by language and ethnicity; such patterns were particularly notable for SLLs, who tended to be involved in advanced coursework and related activities and who rarely interacted with ELLs. Comments related to the social separation of different language groups (its just that its hard to when they have their own group, and Im in mine) suggest that issues of separation are also related to everyday organization of adolescent school-based social life. Additionally, reported out-of-class interactions, such as those between Emily and Javier, where Emily publicly acknowledges Javier in front of other socially connected students, are notable given the importance adolescents place on how they appear in front of others and their belief that imaginary audiences are always watching and observing them (Elkind & Bowen, 1979). Given this context, public declarations such as Emilys could play a role in helping increase ELLs visibility (or sense of visibility) among non-ELL peers in linguistically segregated school contexts.
In relation to the importance of the book-project group structure, Faridas anecdote (among many other similar narratives in the data) suggests that simple integration of cross-linguistic peers in the same classroom may not be sufficient for more meaningful relationships to form. Rather, the intensive and collaborative interaction provided by the book-group project might have played an important role in deepening acquaintanceships, such as hers with Roberto. In this sense LAB provided and proactively promoted a bilingual ecological context for such opportunities through collaborative small-group task design.
Qualitative data also suggest the importance of unique elements of cross-linguistic relationships not captured in quantitative data. For example, students appear to place significant emphasis on talking in the hallways, short greetings or interactions that may be symbolic of social capital development. Although these exchanges are likely indicative of weak ties, they are nonetheless important across segregated groups (Burt, 2004) because they may play a key role in helping students feel integrated into the school environment; in this way, such interactions could be an initial step beyond just sharing the same facilities at the school that moves students toward more integrated experiences.
The results of this study must clearly be interpreted with some caution. Lack of a control group and/or mechanism for assignment to treatment status rules out certain conclusions that can be drawn or generalizations that can be made to larger ELL and SLL populations from this study. Our results also suggest that further studies using a larger participant pool and a variety of study designs and methodologies would be worthwhile in better understanding factors contributing to the development of linguistically and culturally diverse social networks in U.S. secondary schools. While this study helps address the gap in the literature regarding two-way and other integration-focused programs for older students, there are several unaddressed issues related to the social processes through which relationships were built and roles that linguistic ideologies and sociopolitical issues played in the setting, topics which have been addressed, at least partially, in a separate qualitative analysis of interviews, fieldnotes, and audio-recorded group and pair work (Kibler, Salerno, & Hardigree, 2014). One unanswered question is whether or not any carefully structured attempt at integrating linguistically diverse students might have the same impact as LABs two-way dual-language program. While future research would be necessary to explore nuanced responses to this question, we hypothesize that the focus on expertise in both languages might facilitate the development of integrated environments while also offering the extra benefit of supporting bilingualism in U.S. schools. Future work could also usefully explore the durability of social networks after completion of similar programs and the associations among program participation, relationship building, and improved linguistic and academic outcomes.
CONCLUSIONS AND IMPLICATIONS
This analysis provides a window into efforts to change linguistically segregated school environments for the benefit of language majority and minority adolescents through language education programs. Although LABs goals were considerably more modest than those of traditional two-way dual-language programs (which are responsible not just for social development but also all language, literacy, and curricular school learning in both languages), outcomes suggest there may be a role to be played for programs such as LAB that incorporate curricular and instructional approaches that explicitly integrate linguistically diverse students, value English and non-English language skills, and provide opportunities for all students to demonstrate expertise. Many of those Hamilton High students interested in developing cross-linguistic relationships simply did not have a means of doing so within traditional school structures, a situation that research suggests is common across the United States and that speaks to the rather limited linguistic ecologies in which students develop competencies in their second languages. When students are given opportunities and an ecological context that provides affordances (van Lier, 2002) for using language and building these relationships, our research suggests that both ELLs and SLLs may be able to benefit from capital gained in linguistically integrated social networks, in that even weak ties can provide students with new ideas, language use, and perspectives to which they would not have had access in their own linguistically segregated networks, and strong ties arguably can offer even more profound benefits. While programs like LAB do not change the daily lived experiences of students within segregated institutions, they suggest the potential of institutional and curricular bridges (Lin, 2001) that help individuals and peer communities develop empowering ways to participate in their multiple and culturally disparate worlds (Stanton-Salazar, 2004, p. 34) that can serve to facilitate social, linguistic, and academic success.
1. School and student names are pseudonyms.
2. It should be noted that no foreign-language credits were required for graduation with a standard diploma, but at least three were required for college-bound students. Because we recruited from upper-level, non-required Spanish courses, it is likely that these students were college-bound. This was borne out by transcript analysis, below.
3. In network analysis, it is common to also calculate student-level centrality measures. In theory, such a measure would capture the extent to which others report that a given student is very centralthat is, relatively well-connectedwithin the network. In the current paper, we do not focus on the movements of particular students but instead of the formation of ties between students. For this reason, we do not report student-level centrality measures.
4. This is not a traditional quasi-experimental difference-in-differences model because both groups (ELL and SLL) received the treatment (LAB), whereas in a traditional difference-in-differences causal approach, one group is a control. However, strictly speaking we are using the same kind of model and examining whether there is a difference in the pre-/post-differences between ELLs and SLLs, but are not claiming that the difference for SLLs represents a secular trend. In this case, we wish simply to examine and compare mean tie strength in two key groups.
5. Although students were not randomly selected into groups, their group members were not entirely self-selected. In this way, group tie growth would still suggest unique dynamics in groups not present in whole-group settings. We recognize, however, that the lack of random selection does not allow firm conclusions on this point.
6. Student responses in Spanish are presented in italics, with the English translation appearing after a / marking. If students spoke in a mixture of Spanish and English, the former is presented in italics and the latter is kept in regular font.
American Educational Research Association. (2012). Brief of the American educational research association et al. as amici curiae in support of respondents. No 11-345. Retrieved from http://www.aera.net/tabid/13333/Default.aspx
Bearse, C., & de Jong, E. J. (2008). Cultural and linguistic investment: Adolescents in a secondary two-way immersion program. Equity & Excellence in Education, 41(3), 325340. doi:10.1080/1066568080217481
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). Ucinet for Windows [Computer software]. Harvard, MA: Analytic Technologies.
Burt, R. S. (2004). Structural holes and good ideas. American Journal of Sociology, 110(2), 349399. doi:10.1086/421787
Callahan, R. M. (2005). Tracking and high school English learners: Limiting opportunity to learn. American Educational Research Journal, 42, 305328. doi:10.3102/00028312042002305
Carhill-Poza, A. (2011). English language development in context: The peer social networks and interactions of Spanish-speaking adolescent immigrant students (Doctoral dissertation). Retrieved from Proquest, UMI Dissertations Publishing (3454462).
Casciaro, T. (1998). Seeing things clearly. Social Networks, 20, 331351. doi:10.1016/S0378-8733(98)00008-2
Casciaro, T., Carley, K. M., & Krackhardt, D. (1999). Positive affectivity and accuracy in social network perception. Motivation and Emotion, 23(4), 285306.
Coffey, A., & Atkinson, P. (1996). Making sense of qualitative data. Thousand Oaks, CA: Sage Publications.
Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland, J., Mood, A. M., Weinfeld, F. D., & York, R. L. (1966). Equality of educational opportunity. Washington, DC: U.S. Government Printing Office.
Creese, A., & Blackledge, A. (2010). Translanguaging in the bilingual classroom: A pedagogy for learning and teaching? The Modern Language Journal, 94(i), 103-115. doi:10.1111/j.1540-4781.2009.00986.x
de Jong, E., & Bearse, C. I. (2011). The same outcomes for all? High-school students reflection on the two-way immersion program experiences. In D. J. Tedick, D. Christian, & T. W. Fortune (Eds.), Immersion education: Practices, policies, possibilities (pp. 104122). Bristol: Multilingual Matters.
Diem, S., & Brooks, J. (Eds.). (2013). Introduction to the issue on segregation, desegregation, and integration: From history, to policy, to practice. The Teachers College Record, 115(11), 16.
Elkind, D., & Bowen, R. (1979). Imaginary audience behavior in children and adolescents. Developmental Psychology, 15(1), 3844. doi:10.1037/0012-1618.104.22.168
Emerson, R. M., Fretz, R. I., & Shaw, L. L. (2011). Writing ethnographic fieldnotes. (2nd ed.) Chicago: University of Chicago Press.
García, O., & Kleifgan, J. A. (2010). Educating emergent bilinguals: Policies, programs, and practices for English language learners. New York: Teachers College Press.
Gibson, M. A., Gándara, P., & Koyama, J. P. (2004). The role of peers in the schooling of U.S. Mexican youth. In M. A. Gibson, P. C. Gándara, & J. P. Koyama (Eds.), School connections: U.S. Mexican youth, peers, and school achievement (pp. 117). New York: Teachers College Press.
Gifford, B. R., & Valdés, G. (2006). The linguistic isolation of Hispanic students in California's public schools: The challenge of reintegration. Yearbook of the National Society for the Study of Education, 105(2), 125154, doi:10.1111/j.1744-7984.2006.00079.x
Goza, F., & Ryabov, I. (2009). Adolescents educational outcomes: Racial and ethnic variations in peer network importance. Journal of Youth and Adolescence, 38(9), 12641279. doi:10.1007/s10964-009-9418-8
Granovetter, M. (1983). The strength of weak ties, a network theory revisited. Sociological Theory, 1, 201233. doi:10.2307/202051
Hornberger, N. (2002). Multilingual language policies and the continua of biliteracy: An ecological approach. Language Policy, 1, 2751. doi:10.1023/A:1014548611951
Kibler, A., Salerno, A., & Hardigree, C. (2014). More than being in a class: Adolescents ethnolinguistic and social insights in a two-way dual-language program. Language and Education, 28(3), 251-275. doi: 10.1080/09500782.2013.822880
Kramsch, C. (2002). Introduction: How can we tell the dancer from the dance? In C. Kramsch (Ed.), Language acquisition and language socialization: Ecological perspectives (pp. 130). London: Continuum.
Leather, J., & van Dam, J. (2003). Towards an ecology of language acquisition. In J. Leather & J. van Dam (Eds.), Ecology of language acquisition (pp. 130). Norwell, MA: Kluwer.
Lewis-Charp, H., Yu, H. C., & Friedlaender, D. (2004). The influence of intergroup relations on school engagement: Two cases. In M. A. Gibson, P. C. Gándara, & J. P. Koyama (Eds.), School connections: U.S. Mexican youth, peers, and school achievement (pp. 107128). New York: Teachers College Press.
Lin, N. (2001). Social capital: A theory of social structure and action. Cambridge, UK: Cambridge University Press.
Lindholm-Leary, K. (2012). Success and challenges in dual language education. Theory into Practice, 51(4), 256262.
Lindholm-Leary, K., & Hernández, A. (2011). Achievement and language proficiency of Latino students in dual language programmes: Native English speakers, fluent English/previous ELLs, and current ELLs. Journal of Multilingual and Multicultural Development, 32(6), 531545. doi:10.1080/01434632.2011.611596
Marshall, C., & Rossman, G. B. (2011). Designing qualitative research. Thousand Oaks, CA: Sage Publications.
Milroy, L. (1987). Language and social networks. Oxford, UK: Basil Blackwell.
Orfield, G., & Lee, C. (2005). Why segregation matters: Poverty and inequality. Cambridge, MA: Harvard University, The Civil Rights Project.
Orfield, G., Losen, D., Wald, J., & Swanson, C. (2004). Losing our future: How minority youth are being left behind by the graduation rate crisis. Cambridge, MA: The Civil Rights Project at Harvard University. Contributors: Urban Institute, Advocates for Children of New York, and the Civil Society Institute.
Palfreyman, D. (2006). Social context and resources for language learning. System, 34(3), 352370. doi:10.1016/j.system.2006.05.001
Ream, R. K., & Rumberger, R. W. (2008). Student engagement, peer social capital, and school dropout among Mexican American and non-Latino white students. Sociology of Education, 81(2), 109139. doi:10.1177/003804070808100201
Reardon, S. F., Grewal, E. T., Kalogrides, D., & Greenberg, E. (2012). Brown fades: The end of court‐ordered school desegregation and the resegregation of American Public Schools. Journal of Policy Analysis and Management, 31(4), 876–904. doi:10.1002/pam.21649
Rizzuto, T. E., LeDoux, J., & Hatala, J. P. (2009). Its not just what you know, its who you know: Testing a model of the relative importance of social networks to academic performance. Social Psychology of Education, 12(2), 175189. doi:10.1007/s11218-008-9080-0
Rolstad, K. (2005). The big picture: A meta-analysis of program effectiveness research on English Language Learners. Educational Policy, 19(4), 572594. doi:10.1177/0895904805278067
Rumbaut, R. G. (2009). Language graveyard? The evolution of language competencies, preferences and uses among young adult children of immigrants. In T. G. Wiley, J. S. Lee, & R. Rumberger (Eds.), The education of language minority immigrants in the United States (pp. 3571). Bristol: Multilingual Matters.
Rumberger, R., & Palardy, G. (2005). Does segregation still matter? The impact of student composition on academic achievement in high school. Teachers College Record, 107(9), 19992045. doi:10.1111/j.1467-9620.2005.00583.x
Ryabov, I. (2011). Adolescent academic outcomes in school context: Network effects reexamined. Journal of Adolescence, 34(5), 915927. doi:10.1016/j.adolescence.2010.12.004
Ryabov, I., & Van Hook, J. (2007). School segregation and academic achievement among Hispanic children. Social Science Research, 36(2), 767788. doi:10.1016/j.ssresearch.2006.04.002
Salomone, R. C. (2010). True American: Language, identity and the education of immigrant children. Cambridge, MA: Harvard University Press.
Stanton-Salazar, R. D. (2004). Social capital among working-class minority students. In M. A. Gibson, P. C. Gándara, & J. P. Koyama (Eds.), School connections: U.S. Mexican youth, peers, and school achievement (pp. 1838). New York: Teachers College Press.
Suárez-Orozco, C., Pimentel, A., & Martin, M. (2009). The significance of relationships: Academic engagement and achievement among newcomer immigrant youth. Teachers College Record, 111(3), 712749.
Tezanos-Pinto, P., Bratt, C., & Brown, R. (2010). What will the others think? In-group norms as a mediator of the effects of intergroup contact. British Journal of Social Psychology, 49(3), 507523. doi:10.1348/014466609X471020
Titzmann, P. F., & Silbereisen, R. K. (2009). Friendship homophily among ethnic German immigrants: A longitudinal comparison between recent and more experienced immigrant adolescents. Journal of Family Psychology, 23, 301310. doi:10.1037/a0015493
van Lier, L. (2002). An ecological-semiotic perspective on language and linguistics. In C. J. Kramsch (Ed.), Language acquisition and language socialization: Ecological perspectives (pp. 140164). London: Continuum.
van Lier, L. (2004). The ecology and semiotics of language learning: A sociocultural perspective. Boston: Kluwer.
Veronneau, M. H., Vitaro, F., Brendgen, M., Dishion, T. J., & Tremblay, R. E. (2010). Transactional analysis of the reciprocal links between peer experiences and academic achievement from middle childhood to early adolescence. [Research Support, Non-U.S. Government]. Developmental Psychology, 46(4), 773790. doi:10.1037/a0019816
Wiklund, I. (2002). Social networks from a sociolinguistic perspective: The relationship between characteristics of the social networks of bilingual adolescents and their language proficiency. International Journal of the Sociology of Language, 153, 5392. doi:10.1515/ijsl.2002.005
Woolley, M. E., Kol, K. L., & Bowen, G. L. (2009). The social context of school success for Latino middle school students. The Journal of Early Adolescence, 29(1), 4370. doi:10.1177/0272431608324478
Qualitative Interview Protocol
Now we want to know your opinions about LAB, and what the experience was like for you.
Describe your experiences in the LAB program.
Tell me about your experiences writing your bilingual book and reading it to the first- graders.
What have you learned in the LAB program about language? Other people? Yourself?
What have you liked about the LAB program?
What has been challenging about participating in the LAB program?
What goals do you have for learning English/Spanish? In other words, what do you want to be able to do with your English/Spanish in the future?
Prompts for expansion:
Tell me more about ________.
What was _____ like?
How would you describe _______.
Was there a specific time in LAB when you learned that or experienced that?
Can you give me an example?
When did you feel like ____?
What specific memories do you have about ____?
New relationships: this category includes two codes:
Meeting new people (MeetNP): students describe meeting people in LAB whom they had not known before or do not usually see. No reference is made to the types of people they or others are.
Meeting different types of people (MeetDTP): students describe meeting people in LAB who are somehow different than themselves or not the types of people with whom they typically interact and have relationships.
Strengthened acquaintanceships (SAcq): this category itself functions as a code. In these types of excerpts, students describe already knowing another student outside of or prior to LAB but getting to know them better through the program.
Relationships outside LAB: this category includes two codes:
Talking in hallways (TalkH): students describe interactions they have with LAB co-participants in the hallways at school.
Interacting in common school spaces (IntCSS): students describe interactions they have with LAB co-participants in other common school spaces such as the lunchroom or in general spaces (not specifically the hallway).
Interviews: Examples of Each Code, According to Language Designation
Examples were selected to show the range of responses for each code from different students; duplicates from the manuscript were purposely removed in many cases to avoid repetition.
* In some cases, students coded responses did not clearly differentiate between whether or not someone was a new relationship or an existing acquaintanceship. In these cases, we re-read the entire transcript to see if the information was contained elsewhere, and we also referred to students social rankings of their peers.
Interviews: Sample Coded Excerpt
Most student responses to a given interview question could only be applied to a single code, but some responses could be applied to multiple coding categories. The following is an example of how the coding was done within a single response that contained multiple codes:
Abby: I guess that I liked being with other people that I havent normally seen during the day, so get- yeah getting to make new friendships and new friend bases with other people (MeetNP), being on a first-name basis and being able to say hello when I pass them in the hallway, which is really nice (TalkH).
Yesenia: Y, Leah me decía de que ella conocía Honduras. Porque ella había estado ahí. Y, a veces platicábamos de los lugares que ella visitó. Y, yo le decía de lugares que ella necesitaba visitar, como las playas y todo eso (MeetNP). Con Isabella, pues, nos poníamos a hablar, porque a Isabella la conocí antes, mucho más tiempo (SAcq). / Leah told me she knew Honduras because she has been there. And, sometime we talked about the places she visited. And, I told her about places she needed to visit, like the beach and all that (MeetNP). With Isabella, well, we talked because I knew Isabella before, for a much longer time (SAcq).
Once interviews were coded, for each student comment we re-read fieldnotes to find evidence where possible that would support (or refute) it. We then categorized fieldnotes using the appropriate code. (We did not find any contradictory observational evidence relevant to this analysis that refuted students interview comments.) At times, interview comments from different students were supported by a single fieldnote excerpt or a related set of fieldnote excerpts (seen below in the Meet NP example). We then integrated student comments within our formal write-up. An example for selected codes can be found below.