Studies that validate statistical methods for functional magnetic resonance imaging (fMRI) data often use simulated data to ensure that the ground truth is known. However, simulated fMRI data are almost always generated using in-house procedures because a well-accepted simulation method is lacking. In this article we describe the R package neuRosim, which is a collection of data generation functions for neuroimaging data. We will demonstrate the possibilities to generate data from simple time series to complete 4D images and the possibilities for the user to create her own data generation method.