Traditional methods for determining the limit of detection (LOD) and the limit of quantification (LOQ), based on the IUPAC and ACS definitions, often are unsatisfactory. Consequently, a new, simple and statistically correct way of obtaining both the LOD and LOQ values has been derived and compared to commonly used methods. The new Upper Limit Approach, ULA, calculates the upper confidence limit of an individual blank signal using a critical value of the tdistribution and standard error of estimate (residual standard deviation) of regression. The uncertainty of the calibration plot position, its intercept and the mean blank signal are taken also into consideration. The proper choice of calibration model and calibration design also are discussed in detail. An improved derivation of the signal value relevant to the LOQ is based on the use of the same significance level needed for defining the LOD. The concepts developed in this paper have been applied to the determination of cadmium by five common techniques of electrochemical trace analysis (d.c. tast polarography, differential pulse polarography, linear sweep stripping voltammetry, differential pulse stripping voltammetry, and potentiometric stripping analysis) under equivalent (as possible) conditions. The lowest LOD and LOQ values, obtained by the new reliable ULA method, were I . O X ~ O ~ mollL and 3 . 1 ~ 1 0 ~ mol/L Cd, resp., for differential pulse stripping voltammetry and significance level ~ 0 . 0 1 . The relevant LOD and LOQ values for a=0.05 are 6.6 xlO-” mollL and 2.OxlO-’ mol/L Cd, resp. Potentiometric stripping analysis provided similar results.