With only three years of university studies, students at the Swedish Sign Language (STS) interpreter training program are required to develop from novice L2 signers (with none or very limited previous knowledge of STS) to signers capable of interpreting between STS and spoken Swedish. This is a very short time period generally, and in addition, the L2 acquisition primarily occurs through formal classroom instruction due to the limited population of STS signers in the society. Although there is a growing body of research on sign language L2 acquisition internationally, the focus has not been on the teaching. The university STS teachers have therefore (almost) no scientific knowledge to lean on in their instruction. In order to address this lack of scientific knowledge, the longitudinal project UTL2 (Teaching Swedish Sign Language as a second language) was started in 2016, with action research as the main method. The project, in which one researcher collaborated with a team of teachers, is grounded in a desire to change, develop and improve the teachers’ own practice, and to gain new knowledge about effective ways of teaching STS as an L2. Starting with the cohort of students in the program in 2016, the project team examined the use of different languages (STS/spoken Swedish) in the initial instruction through classroom recordings and individual student interviews. In addition, an STS imitation test was developed (SignRepL2) and repeated five times during the first two years of instruction. The knowledge that the team acquired from this first sub-study was discussed in depth and then considered when the 2018 cohort of students started. In addition, a new approach for the instruction was implemented in which the modality-specific training was extended. In this presentation, the project’s preparations, planning, challenges and outcomes will be presented in connection to action research and SLA theories.