An overview of literature data on vapour generation techniques for cadmium and comparison with own experiments by means of several different types of hydride generation–electrothermal atomic absorption spectrometric systems (HG–ETAAS) (batch, semi-batch (SB), continuous-flow (CF) and flow-injection (FI) as well as different gas–liquid separators (GLS) exhibits apparent variations and inconsistency. However, if data for optimal chemical conditions are re-plotted in another coordinates: C HCl (mol l −1) vs. the ratio of reductant-to-acid molar input rates (i.e. millimoles per minute), [BH 4 −]:[H +], much better consistency of data is revealed: more than half of data are clustered around 0.2–0.3 mol l −1 HCl which appears an optimal acidity at moderate BH 4 − concentrations; the tetrahydroborate molar input rates should always be in excess versus the H + molar input rates (1.1 to tenfold); relatively high flow rates of argon purge gas are required (≥120 ml min −1); special attention to the blank control at ng l −1 levels as well as to the construction of gas–liquid separator and vapour transfer lines should be paid. ‘Milder’ conditions for HG could be provided with some of the examined systems and GLSs, thus minimizing reagent consumption, blanks, vigorous reactions, foaming, aerosol production and drift in measurements: e.g. 0.4 mol l −1 HCl—3% m/v NaBH 4 with the semi-batch system and 0.25 mol l −1 HCl—2% m/v NaBH 4 in continuous flow mode. Experimental system is based on the Transversely Heated Graphite Atomizer coupled with flow injection system FIAS 400. Integrated platforms are treated for permanent modification with Zr (110 μg) or W (240 μg) and then with Ir (8 μg). Temperatures of trapping, pyrolysis and atomization are 350, 500 and 1300 °C, respectively. The best overall efficiency of HG, transportation and trapping is 41%. The characteristic mass for peak area measurements is m o=2.8 pg and the limit of detection is 0.002 μg l −1. The long-term stability of characteristic mass (within-day, 8 h) is m o=2.8±0.1 pg (R.S.D. 4.0%, n=8), whereas the corresponding between-day figures (1 mo) are m o=2.8±0.2 pg (R.S.D. 6.6%, n=6). The linear range is 0.002–0.12 μg l −1 with a sample loop of 1.8 ml, being strongly impaired with smaller sample volumes in FI mode. The sample throughput rate is 10 h −1 with the semi-batch system. Applications to real human and bovine urine samples and CRMs of sea water (CASS-3), river water (SLRS-1 and SLRS-3) and urine (SRM 2670) are presented.