To understand the sequential response of the autonomic nervous system to pregnancy, we studied heart rate variability in 23 first trimester, 23 second trimester and 21 third trimester pregnant women. Twenty non-pregnant women were recruited as controls. Time and frequency domain measures of heart rate variability in three recumbent positions were compared. We found that normalized high-frequency power in the supine position increased significantly in the first trimester (42.2 (95% confidence interval (CI) 5.4) nu (normalized unit); P < 0.05) compared with non-pregnant controls (33.0 (6.0) nu), and then decreased progressively in the second (27.3 (6.7) nu) and third (21.8 (6.0) nu; P < 0.05) trimesters. The low-/high-frequency power ratio in the supine position decreased significantly in the first trimester (0.8 (0.3); P < 0.05) compared with that of non-pregnant controls (1.1 (0.3)) and increased progressively in the second (1.5 (0.4)) and third (2.1 (0.8); P < 0.05) trimesters. When the position was changed from the supine to the right lateral decubitus, the percentage change in normalized high-frequency power correlated significantly and negatively with normalized high-frequency power in the supine position in non-pregnant controls (r = -0.56, P = 0.01) and in pregnant women in the first (r = -0.44, P = 0.034), second (r = -0.68, P < 0.001) and third (r = -0.68, P < 0.001) trimesters. These results indicate that autonomic nervous activity shifted towards a lower sympathetic and higher vagal modulation in the first trimester, and changed towards a higher sympathetic and lower vagal modulation in the third trimester as gestational age increased. The balance between the haemodynamic changes of pregnancy and aortocaval compression caused by the enlarging gravid uterus may be responsible for the biphasic changes in autonomic nervous activity during pregnancy.