The relationship between heart rate variability and level of illness was examined. Fifteen patients (10 male and 5 female), gestational ages 25 to 42 weeks, postnatal ages 1 to 42 days, birthweights 545 to 4375 g receiving care in the neonatal intensive care nursery were randomly selected. Data from each infant was transferred from the bedside physiologic monitor to a microcomputer for analysis. A severity of illness index (the Children's Hospital Illness Score [CHILLS]) correlated with heart rate variability. Four patterns of heart rate variability were identified: (1) infants whose CHILLS score indicated that they were cardiovascularly normal demonstrated a modest amount of variability; (2) infants with a CHILLS score indicating a moderate amount of illness had heart rates and heart rate variability greater than the normal infants; (3) three infants exhibited bimodality (period doubling) in their heart rates; each of these infants had a CHILLS score that indicated that they were less ill than the most critically ill patients, but sicker than those moderately ill infants without period doubling; and (4) heart rate decreased in the most critically ill infants identified by the CHILLS, but it remained above the heart rate of a healthy newborn; heart rate variability collapsed below that of a healthy newborn. Our data suggest that the variability of heart rate may increase as an infant becomes sicker. When the infant becomes critically ill and unstable, heart rate variability is less than the normal infant. Nonlinear dynamics theory may be a potential model for fitting the data.