The purpose of the present study was to investigate the retest reliability of event-related functional near-infrared spectroscopy (fNIRS). Therefore, isolated functional activation was evoked in the occipital cortex by a periodic checkerboard stimulation. During a 52-channel fNIRS recording, 12 subjects underwent 60 trials of visual stimulation in two sessions. The retest interval was set to 3 weeks. Linear correlations of the contrast t values supplemented by scatter plots, channel-wise intraclass correlation coefficients (ICC) as well as reproducibility indices for the quantity of activated channels (RQUANTITY) and the location (ROVERLAP) of the detected activation were calculated. The results at the group level showed good reliability in terms of the single measure ICCs (up to 0.84) and excellent reproducibility quantified by RQUANTITY and ROVERLAP (up to 96% of the quantity and the location were reproducible), whereas the results at the single subjects' level were mediocre. Furthermore, the reliability assessed by single measurement ICCs improved if regarded at a cluster level.