Genomic imprinting is an expression of an evolutionary conflict, within the genomes of individual organisms, between genes of maternal and paternal origin. It evolves when a genetically determined action enhances the inclusive fitness of madumnal (maternally derived) genes but reduces the inclusive fitness of padumnal (paternally derived) genes, or vice versa. Such situations arise when the expression of a gene in one individual (the actor) has fitness consequences for other individuals who are asymmetric kin of the actor (Haig Reference Haig1997). (Asymmetric kin are individuals with different probabilities of carrying copies [identical by recent descent] of an actor's madumnal and padumnal alleles. For example, maternal half-sibs are asymmetric kin, related through the actor's mother but not the actor's father, whereas full-sibs are symmetric kin, equally related to the actor via a shared mother and shared father.)

Offspring are usually symmetric kin of parents (a parent's maternal and paternal alleles each has a 50% chance of being transmitted to offspring) but parents are asymmetric kin of offspring (an offspring's maternal alleles are present in its mother, but the offspring's paternal alleles are absent from its mother). Therefore, an offspring's interactions with its parents are predicted to be more internally conflicted than are the parents' interactions with the offspring. If one assumes that interactions with mothers have been more salient than interactions with fathers, then paternally expressed genes should promote behaviors that engage maternal attention and elicit maternal care, whereas maternally expressed genes should favor greater independence of the child. This is the principal source of asymmetric kinship considered by Crespi & Badcock (C&B).

Although an offspring usually has an equal chance of inheriting the madumnal or the padumnal allele at a locus in one of its parents, this symmetry breaks down under inbreeding (Wilkins & Haig Reference Wilkins and Haig2003). Offspring are asymmetric kin of their mother if their father is asymmetric kin of the mother (and vice versa). For example, a child conceived by father-daughter incest is not only its mother's offspring but also her paternal half-sib. Similarly, the child of a woman who marries her father's sister's son is simultaneously its mother's offspring and her patrilateral first cousin once-removed. In both cases, the child would be more likely to carry copies of its mother's padumnal alleles than copies of her madumnal alleles. In general, a woman's madumnal alleles are predicted to favor greater aversion for sexual relations with patrilineal kin (Haig Reference Haig1999a).

Sibling rivalry has probably been intensified in recent human evolution because we have shorter interbirth intervals than our closest relatives but longer periods of juvenile dependence (Kennedy Reference Kennedy2005). Therefore, sibs will often have competed for resources provided by adults. Full-sibs are symmetric kin, whereas half-sibs are asymmetric kin. The key question for understanding the role of imprinting in sibling rivalry is the relative importance of competition with paternal half-sibs (patrisibs) and maternal half-sibs (matrisibs). If the variance of reproductive success is higher among males than among females, then the population will contain more patrisibs than matrisibs. However, offspring tend to maintain closer relations with mothers than with fathers. Therefore, interactions are likely to have been more intense with matrisibs (the products of female infidelity and partner change) than with patrisibs. My expectation is that madumnally expressed genes will tend to promote cooperative interactions among sibs whereas padumnally expressed genes will favor competitive interactions.

Asymmetries of relatedness are also present in a child's interactions with its extended family. Grandparents, aunts, uncles, cousins, nieces, and nephews are all, with rare exceptions, asymmetric kin. Imprinted behaviors might play a role in broader family relations if a child's interactions were predominantly with its mother's kin or its father's kin (or if children evolved specific behaviors for interacting with patrilineal and matrilineal kin). Such asymmetries will have been determined by residence patterns and the stability of pair-bonds (Haig Reference Haig, LeCroy and Moller2000a), whether mothers typically resided with their husband's kin or with their own kin, and whether children followed their mother or stayed with their father after divorce.

I foresee two principal challenges to predicting the effects of imprinted genes on social behaviors from evolutionary first principles. The first challenge (discussed above) will be to understand evolutionarily salient patterns of asymmetric relatedness during human ancestry. The second will be to understand the principal psychological dimensions that have differentially affected matrilineal and patrilineal inclusive fitness, given these asymmetries of relatedness. The latter challenge can be illustrated using the example of a curious “gift” of individuals with Prader-Willi syndrome (PWS). Such individuals perform poorly on most cognitive tests but far outperform individuals from the general population on the assembly of jigsaw puzzles (Dykens Reference Dykens2000). PWS is caused by the absence of expression of genes from padumnal chromosome 15q11–13. This suggests that padumnally expressed genes favor relatively greater development of some psychological attribute that has, as a side-effect, poorer performance on fitting together colored pieces of cardboard. Kinship theory predicts that stronger development of this attribute enhanced individual fitness at a cost to mothers or their families (or reduced individual fitness but conferred a benefit on fathers or their families). The challenge is to understand the nature of the attribute that has been subject to natural selection and how expression of this attribute has differentially affected the fitness of mothers and fathers (or their respective kin).

Both challenges remain to be surmounted in the specific cases of autism and schizophrenia. Infantile onset of autism suggests that the relevant genes may have been subject to selection based on their effects on the mother–infant relationship, whereas the older age of onset of schizophrenia hints that the relevant genes may have been subject to selection for their effects in wider kin networks. The greater challenge is understanding what psychological dimensions have been subject to selection and how variation along these dimensions has affected matrilineal and patrilineal fitness. C&B suggest that autistic features are more pronounced in Angelman syndrome than in Prader-Willi syndrome, whereas Veltman et al. Reference Veltman, Craig and Bolton(2005) conclude the opposite. A detailed study of social behavior in these syndromes is a promising way forward.