Two methods for performing the Thellier experiment on paleointensity determinations are compared. The first method is essentially that of Coe (1967) and the second is a slight modification of it. Prepared samples containing different species of magnetic particles are used, and both methods are always performed on the same sample. The study is designed to accent procedural differences between the two methods. It is shown that only the Coe method yields linear PNRM-PTRM (Partial ThermoRemanent Magnetization) plots for single domain and pseudo-single domain particles, because it maintains the symmetry between high-temperature VRM (Viscous Remanent Magnetization) and spontaneous decay, whereas for the modified Coe method the spontaneous decay is enhanced. For multidomain remanence, even the Coe method yields a nonlinear concave-up PNRM-PTRM plot. Paleointensity determinations employing only the ratio NRM/TRM provide no means of evaluating the reliability of the results.