The mechanism by which heart rate variability (HRV) changes during neonatal illness is not known. One possibility is that reduced HRV is merely a diminished or scaled-down version of normal. Another possibility is that there is a fundamental change in the mechanism underlying HRV, resulting in a change in the ordering of RR intervals. We investigated the nature and extents of order in RR interval time series from 25 Neonatal Intensive Care Unit patients with a spectrum of clinical illness severity and HRV. We measured predictability (deviation of predicted intervals from observed), and regularity (measured as approximate entropy) of RR interval time series showing different degrees of HRV. In RR interval time series where the effects of scaling were removed, we found 1) records showing normal HRV had more order than those showing low HRV; 2) the nature of the order was more like that of a periodic process with frequencies over a large range (time series whose log-log power spectrum had a 1/f distribution) than that of chaotic one (logistic map); and 3) the nature of order did not change greatly as HRV fell. We conclude that neonatal RR interval time series are ordered by periodic processes with frequencies over a large range, and that the extent of order is less during illness when HRV is low.