As data analysis approaches become ever more complex — particularly with data-rich technologies such as functional MRI (fMRI) — the differing analytical choices made by different labs have greater potential to yield different outcomes. Here, a raw fMRI data set was allocated to 70 groups for analysis against nine hypotheses. Although each team chose a different analysis workflow that produced substantial variation in rates of reported significant findings, a meta-analysis that grouped information from all teams showed a consensus in activated brain areas. This study highlights the effect of ‘analytic flexibility’ on conclusions drawn from data and the importance of validation and multiple analyses of the same data set.