Site-over-reference techniques are commonly used to characterize site effects. In these techniques, the choice of reference site significantly affects the amplification estimates. Hence, a clear definition of a standard reference rock site is necessary. The present article tests various standard rock site definitions using the KiK-net Japanese database. Two parameters, the mean shear-wave velocity over the first 30 m (VS30) and the fundamental resonance frequency (f0), are used to define a standard rock site. The variability of the shear-wave velocity profiles and of the estimated rock site response for five different categories, defined according to VS30 and f0 values, are derived and compared. It is shown that the criterion commonly used in building codes establishing a lower limit of VS30 (of about 800 m/sec for rock or soft rock categories) groups a wide variety of shear-wave velocity profiles and a corresponding wide variety in site amplification. A more restrictive, twin criterion, combining (lower) limits on the VS30 and f0 values, significantly reduces the variability in terms of amplification and the amplified frequency band of the rock response. This criterion is thus proposed as a definition of a standard rock site, and its practical applicability is discussed.