We determined the effect of non-uniform specimen thickness underlying an electron probe on specimen thickness measurements by EELS and discussed its effect on quantitation of trace element concentrations. Our results are applicable especially, though not solely, to biological specimens, and show that if the probed area is not uniform, the specimen thickness measured by either the conventional or the fitting method could deviate from the true mean thickness by more than 30%. However, errors associated with these two methods are in the opposite directions and, consequently, the average of the two methods yields the best empirical estimate of the true mean thickness. The error due to non-uniformity of the specimen in the trace element concentration measured in typical cryosections can reach 30%, and its direction (overestimate or underestimate) depends on the methods used for calculating the concentration.