This paper calls for the replication of two task-based language teaching (TBLT) studies, Kim (2012) and Shintani (2012). Kim contributes to the large body of TBLT research on output-based tasks, whereas Shintani initiates research into input-based tasks, a less well researched domain. These studies deserve to be replicated for several reasons. Unlike the bulk of existing TBLT research, they investigate task-based second language (L2) development in addition to performance, and are conducted in actual classroom rather than laboratory contexts. Also, they are well designed studies, including detailed descriptions of the instruments and methodological procedures, making replication feasible. First, this paper provides an overview of the studies. Next, approaches to approximate and conceptual replications are suggested in order to assess the internal and external validity of the original experiments.