An automated TcPO2 data system was tested and used to: 1) quantitate the frequency and duration of hypoxemic and hyperoxemic events experienced during the first three days after birth by 25 very low birth weight infants, and 2) determine whether hypoxemic events were associated with the occurrence of periventricular-intraventricular hemorrhage. During the first three days after birth, infants experienced a median of 41 episodes with a TcPO2 less than 40 torr having a cumulative duration of 37 minutes, and a median of 28 episodes with a TcPO2 greater than 100 torr having a cumulative duration of 14 minutes. Most episodes were less than 30 seconds in duration, but in some infants the duration was longer. There was no association between the occurrence of periventricular-intraventricular hemorrhage and either the frequency or cumulative duration of episodes with a TcPO2 less than 40 torr. A major problem with automated TcPO2 data systems is the inability of currently available systems to recognize inaccurate TcPO2 data. Twenty-four per cent of the summary TcPO2 monitoring data collected in this study was contaminated with unreliable TcPO2 values. Before automated TcPO2 data systems can be recommended for widespread research use or for medicolegal documentation, the problem of data contamination must be resolved.