The aim of this study was to identify factors which might predict nocturnal desaturation (defined as a fall of >4% from awake baseline level for ≥5 min) in normoxic or mildly hypoxic patients with stable COPD [arterial O 2 saturation ( SaO 2) ≥ 91%]. The study was prospective in nature, had full ethical approval and was performed in the Respiratory Department of a city teaching hospital. Thirty-three patients [mean ( sd) age 67·2 (9) years] with stable COPD [mean ( sd) FEV 1 36·8 (11·0) % pred.] were recruited via the respiratory outpatient clinics and through the respiratory wards. The following parameters were measured: daytime arterial blood gases; spirometry; lung volumes (helium dilution); single breath CO transfer factor ( T lCO and KCO); maximum inspiratory (IMP) and expiratory mouth pressures; pulse oximetry ( SpO 2) across a 6-min walk test, and SpO 2 during sleep. Seventeen patients who experienced nocturnal desaturation had significantly lower mean PaO 2 and SaO 2, and higher PaCO 2 values compared to non-desaturators. There was a positive correlation between mean nocturnal SpO 2 and daytime PaO 2, SaO 2, and minimum exercise SpO 2, and a negative correlation between mean nocturnal SpO 2 and PaCO 2, and FRC. Regression analysis revealed that daytime SaO 2 was the only independent predictor of mean nocturnal saturation (accounting for 61% of the variability in the mean nocturnal SpO 2). We observed nocturnal desaturation in all patients with a daytime SaO 2 ≤93% but in no patient with SaO 2 ≥95%. We conclude that daytime SaO 2 can be used to predict nocturnal desaturation in normoxic or mildly hypoxic patients with stable COPD. Nocturnal desaturation is likely in patients with COPD where daytime SaO 2 ≤93%, and unlikely where daytime SaO 2 ≥95%.