The aim of this study was to detect heifers with dystocia using artificial neural networks (ANN). A total of 531 calving records of Holstein-Friesian heifers of Black-and-White strain and 8 diagnostic variables were used. The output variable was the class of calving difficulty: difficult or easy. Perceptrons with one (MLP1) and two (MLP2) hidden layers and radial basis function (RBF) networks were investigated. The root mean square error and the structure of selected ANN (number of neurons in the input, hidden and output layers) were 0.22, 10-4-1; 0.25, 10-17-17-1 and 0.19, 10-25-1 for MLP1, MLP2 and RBF, respectively. The percentage of correctly recognized heifers with difficult and easy calvings and that of correctly diagnosed heifers from both categories for the training and validation sets were approx. 90%. The same values for the test set were 75-83%, 82–88% and 82–86%, respectively. In both cases, no significant differences in these proportions were found. The following variables contributed most to the detection of heifers with dystocia: gestation length, BCS index, CYP19-PvuII and ERα-BglI genotypes and percentage of HF genes in heifer’s genotype.