This paper argues for a broadening of the analytic scope of Interactional Linguistics (IL) to embrace systematic investigation into how grammar grows out of social interaction longitudinally. While IL has amply documented the ways in which grammar structures interaction and emerges locally in real time within and across turns-in-progress, evidence for how social interaction motivates the routinization (or: sedimentation) of grammatical usage patterns over time is scarce due to lack of diachronic interactional data. I review the few existing diachronic and synchronic studies on the issue, and argue that the investigation of second language (L2) interactions opens new avenues for IL research, enabling us to empirically document the over-time emergence of grammar-for-interaction. I then present an analysis of the developmental trajectory of je sais pas ‘I don’t know’ in French L2 interactions across four proficiency levels, showing that the expression progressively routinizes as an interaction-organizational marker in ways that exhibit parallels to diachronic
grammaticalization processes. This demonstrates how the study of L2 interactions can provide developmental evidence for how grammatical patterns emerge and sediment over iterative social encounters as resources for social interaction. The investigation of L2 data hence creates new opportunities for IL research in view of a better understanding of grammar as an outcome of peoples’ acting in the social world.