The aim of the study was to assess how the analysis of different signals recorded during application of automatic continuous positive airway pressure (autoCPAP) devices improves the evaluation of pressure titration in patients with obstructive sleep apnea syndrome (OSAS) naive to treatment. Seventy-two patients underwent nocturnal polysomnography during autoCPAP (Autoset T, ResMed, Sydney, Australia) application. Progressively more complex combinations of signals were analysed in consecutive steps. According to the analysis of oxyhemoglobin saturation (SaO₂) alone, a fixed CPAP level suitable for treatment could not be identified in 3 subjects. When analysis of posture was added, titration was considered unsatisfactory in 1 more subject, due to a short time spent supine. Further, addition of flow and respiratory movements led to consider titration unsatisfactory in 1 more subject. Analysis of all polysomnographic signals demonstrated a not fully reliable titration in 9 subjects: 1 with short sleep duration, 2 without REM sleep, 4 with a short sleep time spent supine, and 3 subjects (already identified by SaO₂) with insufficient correction of respiratory disorders even when a relatively high CPAP was administered. Mask leaks did not hamper titration. CPAP titration by automatic devices alone results in imperfect titration in >10% subjects naive to ventilatory treatment. Only polysomnographic recording ensures titration reliability in all patients. Further research is needed to identify simple and economic methods to reliably start the CPAP treatment.