Reliability and accuracy of the measured ventilatory and lung mechanical parameters of different diagnostic systems (SensorMedics 2600, Med-Science RDS 4500) and respiratory monitoring systems (BI-CORE CP-100, monitor of Babylog 8000) were investigated using a mechanical lung model. The accuracy of pressure and volume signals was measured statically over 6h. The errors of estimated lung mechanic parameters (compliance C, resistance R, time constant T = R.C) were determined for the model parameters Ci = 3.9; 6.4; 10.0; 13.0 ml/kPa and Ri = 4.0; 10.0 kPa/l/s without and with endotracheal tubes (12 Ch, 16 Ch). Altogether 27 parameter combinations were used. The lung mechanic parameters were measured by single occlusion tests. The Babylog 8000 permits only the measurement of the endinspiratory quasistatic compliance C = VT/(Pmax - PEEP). The investigations have shown that -excepted the CP-100 (volume measuring error > 14%)-the accuracy of static measurements was sufficient and in conformity with the allowed tolerances. No significant changes were found over 6h. The mean errors of measured lung mechanic parameters differed extremely in relationship to the time constant of the model. They are for T < 80 ms/T > or = 80 ms 7.4%/-3.5% (SM 2600), -8.4%/-5.8% (CP-100) and -22.0%/-17.3% (Babylog 8000) for compliance, 4.6%/-3.8% (SM 2600) and 189.0%/43.4% (CP-100) for resistance. No reliable measurements of lung mechanics were possible with the RDS 4500 due to software problems. We found, the smaller the time constant the higher the measuring errors of lung mechanic parameters. There are big differences between the investigated devices. Up to now for exact measurements expensive diagnostic systems can not be replaced by respiratory monitors.