Although midazolam is used for sedation of mechanically ventilated newborn babies, this treatment has not been evaluated in a randomised trial. We have done a prospective placebo-controlled study of the effects of midazolam on haemodynamic variables and sedation as judged by a five-item behaviour score. 46 newborn babies on mechanical ventilation for respiratory distress syndrome were randomly assigned to receive midazolam (n = 24) or placebo (n = 22) as a continuous infusion. Doses of midazolam were calculated to obtain plasma concentrations between 200 and 1000 ng/mL within 24 h of starting treatment and to maintain these values throughout the study. Haemodynamic and ventilatory variables were noted every hour, as were complications and possible side-effects of treatment. Mean (SD) duration of inclusion was 78.7 (30.9) h. 1 patient in the treatment group and 7 in the placebo group were withdrawn because of inadequate sedation (p < 0.05). Midazolam gave a significantly better sedative effect than placebo, as estimated by the behaviour score (p < 0.05). Heart rate and blood pressure were reduced by treatment but remained within the normal range for gestational age and there was no effect on ventilatory indices. The incidence of complications was similar in the two groups. No midazolam-related side-effects were noted. Continuous infusion of midazolam at doses adapted to gestational age induces effective sedation in newborn babies on mechanical ventilation, with positive effects on haemodynamic variables. The course of the respiratory distress syndrome was not influenced by this treatment. Midazolam was given over only a few days and the limited effects on heart rate and blood pressure that we report should not encourage long-term administration.