Inasmuch as spinal taps in preterm infants are frequently accompanied by clinical deterioration, the optimal position for this procedure was investigated. Three positions were each randomly assigned for five minutes to 17 healthy preterm infants without a spinal tap actually being performed: (1) lateral recumbent with full flexion (flexed position), (2) lateral recumbent with partial neck extension (extended position), and (3) sitting with head support and spine flexion (upright position). Transcutaneous PO2 and PCO2 were monitored in all infants, minute ventilation (VI) in seven, and heart rate and blood pressure in ten infants. Mean transcutaneous PO2 decreased in each of the three positions. This decrease was significantly greater in the flexed (28 +/- 8 mm Hg) as compared with the extended (18 +/- 8 mm Hg, P less than .001) and upright (15 +/- 11 mm Hg, P less than .001) positions. Mean transcutaneous PCO2 increased only in the flexed position (3 +/- 4 mm Hg, P less than .005) and levels were still elevated five minutes after that position had been discontinued. The consistent decrease in transcutaneous PO2 was accompanied by a variable effect of positioning on VI and there were no episodes of airway obstruction or apnea greater than 10 seconds. Heart rate increased in each position whereas blood pressure remained unchanged. These data suggest that although hypoventilation may contribute to the observed decrease in transcutaneous PO2, ventilation/perfusion imbalance appears to be the major mechanism. As spinal taps performed in the widely accepted flexed position carry the greatest risk of potential morbidity, it is recommended that this position be modified with neck extension or that spinal taps be performed in the upright position.