OBJECTIVE: In extremely premature neonates, data concerning the normal baseline variability of near-infrared spectroscopy (NIRS)-derived regional oxygen saturation (rSO 2 ) are lacking. We sought to determine: 1) the quiescent variability of cerebral, renal, and splanchnic rSO 2 in clinically stable, undisturbed very low birth weight neonates and 2) the effects of different data averaging epochs on site-specific variability. STUDY DESIGN: In this prospective, observational study, neonates between 500 and 1250 g underwent seven days of continuous, real-time cerebral, renal, and splanchnic NIRS monitoring starting within the first seventy-two postnatal hours. Demographic, cardiopulmonary, bedside care, and rSO 2 data were collected. rSO 2 variability was analyzed utilizing data from quiescent periods identified using pre-specified stability criteria. Between-and within-monitoring site comparisons of data averaging methods were made utilizing ANOVA. RESULT: Twenty-four subjects (GA 27 ± 0.3 wk, birth weight 988 ± 34 g; mean ± SEM) were monitored. Coefficients of variation (CoVar = SD/mean) were calculated for each monitoring site using varied data averaging epochs. CoVar was lowest for cerebral, intermediate for renal, and highest for splanchnic rSO 2 (P < 0.01). For renal and splanchnic sites, shorter epochs (5-and 15-min) resulted in significantly smaller CoVars [P < 0.01 and P < 0.05, respectively]. Splanchnic variability was highly dependent on epoch length, ranging from 16% over 5 min to 23% over 60 min. CONCLUSION: 1) rSO 2 variability differs significantly between monitoring sites and 2) shorter data sampling epochs decrease rSO 2 variability. These observations may assist clinicians in operationally defining minimally significant departures to enable medical decision making utilizing this monitoring technique.