The large number of different breath instruments in use and uncertainty about calibration procedures and also the fact breath sampling was done manually, probably explains the large residual SD of 0.40 g/kg, which is an indication of random variation. For comparison, the residual SD was 0.15 g/kg when an evidential breath analyser (Intoxilyzer 5000S) was used to test drunken drivers and when the calibration was controlled with every subject test