Many long-lived birds form long-term pair bonds, but they sometimes divorce. Until now most suggestions concerning factors relevant to the stability of the pair bond in birds were based on correlational data, which do not allow conclusions about causal relationships. Here, the probability of divorce in domestic pigeons (Columba livia) was tested experimentally, and separately for both male and female. Breeding success and pair bond duration of pairs was manipulated. After separation from the mate for up to 60 days the subject (female or male) was allowed to court to a new mate for 4 hours. Then the subject was given a choice between the old and the new mate presented simultaneously on either side of the experimental cage. A further possibility to choose between the two mates was given in later stage in the familiar aviary. Female subjects without breeding success showed little courtship towards the old male, especially when they had nest-cooed a lot to the new male during the four hours exposure. In the latter cases they chose the new male in the experimental choice situation. The preference for the old mate in the choice situation by females with breeding success was positively related to the duration of the previous pair bond. All females that had been paired for at least 200 days (N = 11), chose their old males. On the other hand, 86% of the females that had been paired for less than 152 days, chose the new male. Forty percent of the male subjects with breeding success preferred the new female in the experimental choice situation. This preference was especially shown by males that had performed frequent nest-cooing during the four hours exposure period. In the aviary, however, all males with or without breeding success reoccupied their nest at once and started nest-cooing and finally courted the female which was able to defend access to the nest against other females. This was always the male's former mate. Males preferred to occupy a nest over courting with any particular female. The results are discussed in relation to existing hypotheses of divorce.