This study examines signal availability in fetal electrocardiogram (FECG) beat-to-beat acquisition and the accuracy of fetal heart rate variability (HRV) analysis in the clinical setting using a commercially available FECG monitor. Signal availability was examined in 130 FECG recordings of 0.3-17.5 h duration collected in 63 fetuses (25th-42nd week of gestation) under uncontrolled conditions. Identification of R-peaks demonstrated a signal loss of 30% +/- 24% with 3.6 +/- 1.7 signal gaps per minute. Median duration of the gaps within a recording was 1.8 +/- 0.2 s. Per hour of recording, 1.8 +/- 2.1 episodes of 5 min of uninterrupted data were found. Signal availability improved with gestational age and was poorer in women with high body-mass index. Fetal HRV between weeks 36-42 was examined on the basis of 5 min RR-interval episodes obtained under controlled quiet conditions in 55 FECG compared to 46 high quality fetal magnetocardiograms. There were no differences in RR-interval duration, its standard deviation and low frequency power. However, various measures of short-term HRV were significantly higher in the FECG data: root mean square of successive differences (10.0 +/- 1.8 versus 6.6 +/- 3.0 ms, p < 0.001, high frequency spectral power (24 +/- 12 versus 13 +/- 13 ms(2), p < 0.001) and approximate entropy (0.86 +/- 0.16 versus 0.73 +/- 0.24, p = 0.007). We conclude that, in spite of considerable signal loss, FECG recordings can accurately estimate heart rate and its overall variance. However, measures that quantify short-term beat-to-beat HRV will be compromised due to possible recurring inappropriate detection of single R-peaks.