In order to verify the effect of almitrine bismesylate on episodes of nocturnal oxygen desaturation, 10 patients suffering from chronic obstructive pulmonary disease had polygraphic EEG night sleep recordings with simultaneous measurement of O2 saturation. The subjects presented during nocturnal sleep a quantity of desaturation of at least 2,400 sec. X p. cent (this quantity being represented by the surface of the SaO2 curve under the desaturation level defined as the waking SaO2 less 4 p. cent) and had been included in a matched experimental protocol, double blind, almitrine bismesylate versus placebo. After administration of either placebo or almitrine bismesylate for 14 days (100 mg per day in two doses) a control night sleep recording was performed. The effect of treatment was assessed using an analysis of variance for 2 or 3 factors, completed by a Student t test in case of significant interaction between the factors. It appears that almitrine bismesylate produces: a non significant decrease in total sleep, slow wave sleep and REM sleep time; a significantly earlier awakening (p less than 0.05); a significant reduction of the total duration of desaturation episodes (p less than 0.05) without any correlation with the variations of total sleep time; a non significant decrease in the quantity of desaturation; a significant increase in PaO2 (p less than 0.001) and in SaO2 (p less than 0.01) measured two hours after the awakening. However there was no evidence of any significant change in mean SaO2 and lowest SaO2 during the night. Finally the variability of the effect of almitrine bismesylate during day or night time remains unexplained.