One might think that the question whether a female Peromyscus polionotus has sexual relations with more than one partner is nobody's affair but her own. Hurst1, however, argues that the coexistence of partner fidelity with genomic imprinting in P. polionotus2 adds to evidence contradicting the 'conflict hypothesis', which states genomic imprinting has evolved because of the conflicting interests of maternal and paternal genomes3,4.

P. polionotus is not strictly monogamous, but Hurst questions whether the low rate of partner exchange is sufficient to maintain imprinting given the cost of increased exposure to deleterious recessives when one allele is silent. Neither the selective cost of deleterious recessives nor the rate of partner change is known with precision, but a rough comparison suggests that paternal turnover is sufficient for the selective maintenance of genomic imprinting. At mutation-selection equilibrium, there is one selective death for each deleterious mutation5. Therefore, the selective force favouring the loss of imprinting is of the same order of magnitude as the mutation rate. By contrast, field data suggest that female P. polionotus frequently change partners (although not as often as Peromyscus maniculatus). Foltz6 collected 61 pairs of consecutive litters of P. polionotus, and estimated that 12 subsequent litters (20%) had a different father from the mother's previous litter.

Inactivating mutations of several imprinted loci have phenotypes that Hurst believes contradict the conflict hypothesis1. I would prefer to wait until more is known about the normal functions of some of these loci before judging one way or the other, but will briefly comment on the case of paternally expressed Mest. Newborn Mest-deficient mice are growth retarded—consistent with predictions of the hypothesis—but Mest-deficient mothers neglect their young, which Hurst remarks as being "a behaviour that cannot obviously be explained by the conflict hypothesis." One may envisage, however, that if Mest has pleiotropic effects on growth and behaviour7, imprinting might be maintained because of the gene's effects on fetal growth, with the imprinting of maternal behaviour a side effect. Nevertheless, one should not reject without test the possibility that natural selection has favoured imprinting of genes that affect maternal behaviour. Mice often form communal nests, with a preference for female relatives8,9,10. A mother's paternal genome would be selected to provide greater care for the mother's own offspring than would the mother's maternal genome, if by so doing the reproductive values of her own mother, her maternal half-sibs, or other maternal-side relatives were reduced3.