Model-based statistical analysis of functional magnetic resonance imaging (fMRI) data relies on the general linear model and statistical hypothesis testing. Due to the large number of intracranial voxels, it is important to deal with the multiple comparisons problem. Many fMRI analysis tools utilize Gaussian random field theory to obtain a more sensitive thresholding; this typically involves Gaussian smoothing as a preprocessing step. Wavelet-based statistical parametric mapping (WSPM) is an alternative method to obtain parametric maps from non-smoothed data. It relies on adaptive thresholding of the parametric maps in the wavelet domain, followed by voxel-wise statistical testing. The procedure is conservative; it uses Bonferroni correction for strong type I error control. Yet, its sensitivity is close to SPM's due to the excellent denoising properties of the wavelet transform. Here, we adapt the false discovery rate (FDR) principle to the WSPM framework. Although explicit p-values cannot be obtained, we show that it is possible to retrieve the FDR threshold by a simple iterative scheme. We then validate the approach with an event-related visual stimulation task. Our results show better sensitivity with preservation of spatial resolution; i.e., activation clusters align well with the gray matter structures in the visual cortex. The spatial resolution of the activation maps is even high enough to easily identify a voxel that is very likely to be caused by the draining-vein effect.