Thirty-five 2-h recordings of fetal heart rate and fetal movements, made at 38-39 weeks of gestational age, have been analyzed in a partly automated procedure. Involved were 17 nulliparous and 19 multiparous women. To minimize subjectivity the classification of the heart rate was performed in a Delphi group opinion procedure consisting of three rounds. In the third round, three independent investigators classified 89.1% of the total tracing time as pattern A through D. The investigators could not classify 6.1% of the tracing time in any of the available categories. The percentages of coincidence of state parameters did not differ significantly between the fetuses of nulli- and multiparous women. Fifty percent of the fetuses showed true behavioural states, both in the nulli and the multiparous women. The percentage of time spent in state 1F was higher in the multiparous group (P less than 0.05). The other percentages of states did not differ significantly, neither did the duration of the enclosed epochs. The advantages and disadvantages of the automated assignment of fetal behavioural states are discussed.