Bilingual-education scholars have argued for the need for longitudinal, interventionist studies (e.g. Lin, 2009) and for mixed research methods in order to understand the multi-causality nature of schools and classrooms (Baker & Wright, 2017). Prior research has revealed challenges in classroom/school contexts in combining methodological rigor with participant fluidity (Dörnyei, 2007), which adds to the complexity of doing school-based research. In this presentation, we describe an ongoing, four-year, longitudinal mixed-methods collaborative research venture involving four teacher trainers/researchers and six teachers at four multilingual secondary schools in Sweden. We also present results, and we conclude by discussing reciprocal benefits and challenges as well as the balance between research rigor and participant fluidity in this project.

The study addresses issues of social justice in English-as-an-additional language (EAL) classrooms in multilingual schools, and the need for research-based teaching/learning materials of vocabulary in EAL. In the overall design, we were guided by the ethnographic discourse-analytic framework Nexus Analysis (NA) (Scollon & Scollon, 2004; Hult, 2017) and its three phases of engaging, navigating and changing the nexus of practice, in our case multilingual EAL classrooms. We engaged EAL classrooms by initiating collaboration with school leaders and teachers; we then navigated the classrooms by conducting observation and interviews, and by video-recording EAL lessons. Finally, change was introduced by us (researchers/teacher trainers) teaching a three-week intervention in six intact year-9 classes, systematically changing the de facto classroom language policy (Johnson, 2014) by inviting students to draw on different languages when engaging with new English vocabulary. Classroom language practices were analysed by applying the NA discourses ‘historical body’, ‘discourses in place’ and ‘interaction order’ to observation, interview and photography data. Intervention data of students’ learning of vocabulary was analysed quantitatively, using descriptive and inferential statistics of repeated-measures design vocabulary learning data, involving pre-tests, immediate and delayed post-tests.