It is well known that some minerals give underestimated luminescence ages due to anomalous fading. The anomalous fading follows a logarithmic decay law characterized by its slope, the socalled fading rate or g-value. Using the fading rate, Huntley and Lamothe (2001) suggested some correction for the fading underestimation of young samples (<40-50 ka). For polymineral fine grains, we observe a fading rate of 0-4%/decade for TL and BL-OSL and 4-6%/decade for IR-OSL. Extending the laboratory observation to archaeological age, the underestimation on the age for 10 ka is estimated to a mean of 5% for TL, 10% for BL-OSL and 45% for IR-OSL. Due to the non-linearity of the Huntley and Lamothe's fading correction, the contribution of the fading to the total uncertainty is estimated by a Monte-Carlo simulation. The inference on dating shows that the uncertainty on the anomalous fading can be a significant term of the combined uncertainty on the age, even for low fading rates.