We determined the effects of betamethasone on the fetal sheep electrocorticogram (ECoG) using linear (power spectral) and non-linear analysis. For non-linear analysis we used an algorithm based on the Wolf algorithm for the estimation of the leading Lyapunov exponent which calculates a prediction error based on the course of the time series in the phase space. A high prediction error stands for low predictibility or low regularity and vice versa. After 48 h of baseline recordings, vehicle (n = 6) or betamethasone (n = 7) at 10 microg h(-1) was infused over 48 h to the sheep fetus at 128 days gestational age (0.87 of gestation). ECoG spectral analysis revealed no difference in power spectrum between vehicle- and betamethasone-treated fetuses. The prediction error of the ECoG during REM sleep was higher than during non-REM or quiet sleep in both groups (P < 0.0001) revealing lower causality of brain activity during REM sleep. During REM sleep, prediction error significantly decreased 18-24 h after onset of betamethasone treatment (P < 0.05) and returned to baseline values within the following 24 h of continued betamethasone treatment. No ECoG changes were found during quiet sleep. Non-linear ECoG changes during metabolically active REM sleep accompanied the previously described decrease in cerebral blood flow. These results suggest that betamethasone in doses used in perinatal medicine acutely alters complex neuronal activity.