Before the 1960s newborn infants with severe lung disease, usually due to respiratory distress syndrome (RDS), had a very high mortality rate. Standard treatment consisted of supportive measures including supplemental oxygen and correction of metabolic acidosis. Mechanical ventilation (MV) was introduced in the 1960s to correct hypoxaemia and respiratory acidosis in infants who were likely to die. MV is now standard treatment for infants with severe RDS but the degree to which this made a contribution to the outcome of such infants compared with standard neonatal care, is uncertain. To evaluate the effects of the use of MV compared with no MV on mortality and morbidity in newborn infants with severe respiratory failure due to pulmonary disease. Searches were last updated in March 2002 on the Cochrane Controlled Trials Register (Cochrane Library Issue 1, 2002), MEDLINE from 1966 and EMBASE from 1980. In order to detect trials that may not have been published in full, searches carried out of the Oxford Database of Perinatal Trials and for abstracts published by the Society for Pediatric Research (1967 to 2001) and the European Society for Pediatric Research 1970 to 1977. Experts were consulted with emphasis on those who were in active neonatal practice in the 1960s and 1970s when the majority of these trials were likely to have been done. Randomised or quasi-randomised controlled trials in newborn infants with respiratory failure due to pulmonary disease evaluating the use of MV versus standard neonatal care without MV. The standard methods of the Cochrane Collaboration and its Neonatal Review Group were used. Two authors independently assessed eligibility, methodological quality of each trial and extracted the data. Additional information was obtained from all trial authors on methodology or data. The data were analysed using relative risk and risk difference and their 95% confidence intervals. A fixed effect model was used for meta-analyses. The five trials reported on a total of 359 infants with RDS. In one study there is a higher neonatal mortality in the mechanical ventilation group [7/10 vs 1/10; RR 7.00 (1.04, 46.95)]. Overall any reported mortality is less frequent in the mechanical ventilation group with the upper 95% confidence limit on 1.00 [summary RR 0.86 (0.74, 1.00), RD -0.10 (-0.20, -0.01), NNT 10 (5, 100)]. In infants with a birth weight of 1 - 2 kg, no significant difference in mortality is found [summary RR for two trials 0.86 (0.70, 1.07)]. In infants with a birth weight of more than 2 kg, one study reports a significant reduction in mortality in the MV group compared with control [RR 0.67 (0.51, 0.86)]; overall for this birth weight group there is a significant reduction in mortality with MV in the two trials [summary RR 0.67 (0.52, 0.87), RD -0.27 (-0.45, -0.10), NNT 4 (2, 10)]. Any IVH at autopsy is not significantly different between the groups in any study or overall in four studies reporting on 202 infants who had an autopsy. Pneumothorax was reported in two studies of 275 infants and there is a non-significant trend towards an increase in the mechanical ventilation group [summary RR 2.75 (0.72, 10.45)]. When MV was introduced in the 1960s to treat infants with severe respiratory failure due to pulmonary disease, trials showed an overall reduction in mortality which was most marked in infants born with a birthweight of more than 2 kg. This review does not provide information to evaluate the relative benefits or harms of MV in the setting of modern perinatal care. In settings without modern neonatal care, the introduction of MV should ideally be evaluated in randomised controlled trials for its relative benefits, harms and costs.