One of the most important foundations of luminescence dating is the assumption that the growth of the luminescence signal in nature can be reproduced under laboratory conditions by performing irradiations with a calibrated beta or gamma source. When optically stimulated luminescence (OSL) of quartz with a dominant fast component is measured using the single aliquot regenerative dose (SAR) protocol, laboratory dose response curves that display continuing growth at high doses are increasingly reported in literature. In this study we investigate fine (4–11 μm) and coarse (63–90 μm) quartz extracted from 25 samples taken from L1, S1 and L2 units from the loess-palaeosol section at Costineşti in Romania. Our results indicate that the growth of the OSL signal in nature does not correspond to the laboratory generated laboratory dose response curve. The growth of the signal in nature is consistent with a single saturating exponential function, with the signal of coarse grains starting to saturate at 100–200 Gy, and for fine grains at 200–300 Gy, respectively. Laboratory dose response curves continue to grow for high doses (>300 Gy) for both quartz fractions. The differences observed between the natural and the laboratory dose response for the two quartz fractions are believed to be a cause for the different chronologies previously reported using the two grain sizes of quartz on Romanian loess. In addition, we have applied the single aliquot regeneration and added dose (SARA) procedure to both fine and coarse grains from the youngest sample. Our findings question the reliability of obtaining high equivalent doses for quartz samples displaying laboratory dose response curves obtained by the SAR protocol for which a single saturating exponential model does not describe the data.