In non-polluted natural waters, trace metals have generally to be determined near the detection limit of electrothermal atomic absorption spectrometry (ETAAS) and therefore the experimental conditions must be optimized. The detection limit in ETAAS depends strongly on numerous experimental parameters and its optimization is time consuming because, according to IUPAC recommendations, 20 repeated blank measurements per experimental parameter must be performed. In this work, for lead the influence of different experimental parameters on the instrumental detection limit together with the relatively good agreement between experimental and a priori calculated detection limits was demonstrated. The effects of spectroscopic and electrothermal parameters were elucidated from the calculation of detection limits. Under optimized conditions, the detection limit obtained for Pb at 283.3 nm was half that at 217 nm; a 30% improvement was obtained on replacing the hollow-cathode lamp with an electrodeless discharge lamp. The calculation of the detection limit allowed a comparison of atomizers of various geometries and confirmed that high atomization temperatures can be used with the Perkin-Elmer Model 4100ZL spectrometer without degrading the detection limit for Pb. The work on lead also demonstrated that optimizing the integration window by determining the start and end of integration with a computer after smoothing the data significantly improved the detection limit. Under the experimental conditions adopted, using a moving sum of about 20 points led to the lowest detection limit. With a relatively important background absorbance signal, the detection limit for the same dosing volume was improved by using the multiple-injection mode with summation of the resulting atomization signals instead of using the single injection mode.