Several methods of linear regression reported in the literature for use in geochonology are discussed in order to guide geochronologists in the selection of an adequate model and also to distinguish among isochrons, errorchrons and pseudoisochrons. For Rb/Sr isochron methods, at least 6 models or statistical treatments are available, each one with its peculiarities. Based in two examples, we show that for cases when the experimental data points scatter within limits determined by the experimental errors, and using or not the weighted regression treatment, the resultant parameters are essentially concordant. The estimates of associated error, however especially for initial ratios, show variations by up to a factor of 7, depending upon the statistical procedure. In view of this fact and on the basis of simulated test data, we propose not only more adequate models, but also general criteria that must be observed in the selection of samples in order to obtain an isochron. If the data to be regressed show a scatter in excess of experimental error for geological reasons (open system), four regression models are available to treat the parameters defined in these cases as errorchrons. These models are briefly discussed in terms of their applications. In the Pb/U methods, in which the usual regression treatment is applied to obtain the best-fit line, some additional peculiarities are pointed out
