Abstract: In use of a UV gas discharge tube (such as used in flame monitoring apparatus), an electric field is periodically applied in the tube, each application of the field being followed by an ‘off’ period in which the field is removed. During this process, the mean value of the statistical lag Ts is measured over a predetermined time duration (the statistical lag is the time lag after each application of the electric field to the tube before conduction (if any) takes place). If the statistical lag lies within region I, the flame is judged to be present. If the statistical lag lies in region II, the flame is judged to be off (and a warning may be signalled). If the statistical lag lies in region III, a fault in the tube is signalled. This may be a “field emission” fault whereby free electrons are generated by the applied electric field, without the presence of UV radiation or it may be a “multiple counting” fault.