Temporal signal-to-noise ratio (tSNR) is commonly used as a quality assurance measure of fMRI data. The aim of this study was to investigate how reproducibly tSNR maps could be collected in-vivo. We collected 16 fMRI time series in several sessions on two different days and calculated tSNR maps for each. The variability was evaluated over all runs, runs within the same session, runs between sessions and runs between days. Reproducibility turned out to be very poor, which precludes using tSNR as an outcome measure for testing variables that are expected to provide modest but otherwise important improvements to data quality.