Advanced analytical tools play a central role in human brain mapping research. The validation of these tools, however, is particularly challenging in the absence of a ground truth. Sound method papers generally include some benchmark evaluations on simulated data, where the ground truth is known and different scenarios can be tested. If these simulations are based on simplistic assumptions, as is most often the case, such experiment holds more as a sanity check than an actual demonstration of validity. Recently, fMRI simulations with realistic properties have started to emerge in the context of method validation. The results have sometimes been very surprising, challenging common practice in fMRI data analysis, such as the inflated family-wise error in cluster-based inference implemented in many popular packages (FSL, SPM, AFNI). In this symposium, we will present a number of validation works, covering established methods (falsediscovery rate and cluster-based inference in group general linear models) as well as emerging techniques (artifact reduction using independent component analysis). The simulation models themselves will cover a range of techniques (resampling of real data, linear mixture of real spatial component, multimodal computational model of brain connectivity). Importantly, each speaker will present works based on public software packages that can be used to implement these solutions.