This study consisted of the following aspects : a) calibration b) reference methods c) influence of moisturizers. a) Calibration and on-line verification Two calibration methods can be envisaged : 1) On line. Calibration is difficult in view of the large quantities of tobacco required to obtain a sufficient range of differences from tobacco with constant moisture content. 2) Static determination. Calibration is easier, but the use and verification of this method carry risks of deviation during the determination of a flow of tobacco. b) Reference methods in the laboratory, the moisture content of tobacco is determined by various methods, which differ from continuous monitoring. These methods measure either the"total"water content (Karl Fischer, azeotropic, GC) or a loss of volatile substances, including water (oven method, for example). The on-line apparatus determines the surface moisture of the strands, whereas the reference methods determine a"total"moisture conte nt. There are therefore two sources of differences : _ arising from the reference method. _ arising from the distribution of the moisture within the strands. c) Influence of moisturizers. The moisturizers are partially taken into account during determinations of the loss of volatile substances, by an increase in the reference moisture values. In on-line methods involving IR determination, the moisturizers are not taken into account and may even act as screens. This study has not resolved all these problems, but is intended to bring the problems encountered in using an on-line measurement system into clearer focus.