Flexible adaptation is critical for human goal-directed behavior, as the current context we find ourselves in can change frequently and rapidly. One key process supporting this ability is the flexible neural representation of current tasks and goals. While past research clearly demonstrated that frontal and parietal brain regions contain task representations, findings on whether and how these representations change across contexts remain inconsistent. This can, in large parts, be explained by a relatively small number of often underpowered fMRI studies and very few replication attempts, leading to low replicability in this field. In the proposed project, my aim is to directly tackle the replication issue by first performing an fMRI mega-analysis, i.e. a pooling of previously published results on the single-subject level. This will allow me to identify whether and how task representations change across contexts using much higher sample sizes and significantly increased experimental power. In a second step, I will take the main findings of this mega-analysis, and replicate them in a novel data-set, using a pre-registered, fully powered fMRI study design. By putting the replication crisis at the core of the proposed research, I believe this project will have a substantial impact, on task coding research and cognitive neuroscience more broadly. It will allow future research and theory-building to rest on a much firmer, reliable empirical basis.