Errors in dose assessment introduced by the use of the ''linear part'' of a saturating dose response curve

Linear fitting is still often applied when determining past radiation dose values in luminescence and ESR dating. This approach has been justified by the apparent linearity of the dose response data and better internal consistency of the data sets when applying linear fitting compared to exponential fitting. This study demonstrates that if the dose response follows a single saturating function, linear fitting causes very large systematic uncertainties, even when applied to sections of the dose response curve that show linear regression coefficients of 1. Although under certain circumstances linear fitting can indeed provide smaller random errors (i.e. a better internal consistency) than exponential fitting, the results of this study strongly suggest abandoning linear fitting altogether.