Fetal heart rate variability reflects fetal development and is affected by pathological conditions. The aim of this study was to examine fetal heart period irregularity by focussing on nonlinear dynamical components. We recorded 165 fetal magnetocardiograms in 39 healthy pregnancies between the 16th and 41st week. The irregularity of the RR time series of each recording was quantified on the basis of the approximate entropy (ApEn). In order to estimate the nonlinear component, a surrogate time series was generated for each time series and ApEn was again calculated. Next, the purely dynamical aspect of the series was examined using a binary representation reflecting increase or decrease in RR interval of the original time series. ApEn of the original time series increased during pregnancy with a significant dependence on gestational age (p& #60>;0.0005). The ApEn values of the surrogate data increased similarly although the power of the relationship to week of gestation was weaker (roriginal=0.68, rsurrogate=0.46). Also the values of the surrogate series were generally higher, the mean difference being 0.24±0.24 (p& #60>;0.0005). ApEn of the binary series revealed no dependency on gestational age but again the values for the surrogate series were higher than those for the original series. The increase in irregularity of RR interval time series during pregnancy can be attributed in part to an irregular, nonlinear temporal structure and is not solely due to linear autocorrelations. Reducing information by constructing time series using binary symbolization which ignores the absolute beat durations resulted in a loss of dependency on gestational age but a retention of nonlinearity. The ability to quantify these processes promises to aid in prenatal risk stratification.