Electrophysiological measurement of uterine contractions, referred to as electrohysterogram (EHG), is potentially more informative than methods currently used during pregnancy for timely recognition of complications such as preterm labor. Unfortunately, EHG measurement and interpretation remain challenging. Recently, some attention has been dedicated to the analysis of the EHG propagation, which is hypothetically predictive of the delivery time. This hypothesis, though physiologically reasonable, has not been investigated yet. A dedicated maximum likelihood (ML) method has been proposed and validated for identifying the conduction velocity vector of single EHG spikes. This validated ML method is here employed for comparing the conduction velocity vector in two groups of pregnant women with uterine contractions that were prospectively classified as productive or unproductive contractions. The estimated conduction velocity vector showed significant differences in the two groups. The spikes extracted from those contractions eventually classified as unproductive showed a significantly lower conduction velocity amplitude (CV = 4.89 ± 1.19 cm.s(-1) vs CV = 8.63 ± 2.92 cm.s(-1)) and a higher occurrence of upward propagation relative to productive contractions. These results suggest that productive and unproductive uterine contractions are associated to significantly different properties of the conduction velocity vector, which is likely to be proven fundamental in predicting preterm delivery.