As the pantomime season approaches, would-be players around the country will be jostling for their coveted roles. Even for Cinderella's Wicked Stepmother, there will be no lack of eager thespians willing to play the part.

In real life, however, people tend to approach step-parenthood with much less enthusiasm; whether they are wicked, saintly, or just ordinary human beings who sometimes find it hard to behave as well as they would like. Martin Daly and Margo Wilson, who recently spoke on 'The Truth About Cinderella' at one of the London School of Economics' Darwin Seminars, point out the limited usefulness of referring to parenthood as a 'role'.

As social scientists, and specifically as evolutionary psychologists, Daly and Wilson want to view family relationships within the framework of theory. "People talk about role theory, says Daly, but "it seems to be nothing but a metaphor, that it's theatre. You've got to know the part. There's some truth to that, clearly; you've got to know what dads are supposed to do to go about being a dad.

"At the same time, it drives passions out of human behaviour, if you take it too seriously. It leaves out of consideration the fact that people resent being obliged to play certain roles, and enthusiastically embrace certain others."

Thinking within a framework based on the idea of natural selection, Daly and Wilson hypothesised that the parental psyche must have evolved to embrace children and their rearing. But readiness to play this role would depend on kinship; on shared genes. As the role is an exceptionally demanding one, which has to be sustained even when the object of parental attention is not giving anything in return, Daly and Wilson's prediction was that step-parents are less likely to maintain this commitment. At its farthest extreme, this tendency to lapse would be reflected in the statistics on the killing of children.

At this point, sceptics may be inclined to ask why we need Darwin to guide us to this possibility. After all, 'Cinderella' and other folktales featuring wicked step-parents are an expression of a folk understanding that stepfamilies are not the same as genetic ones, and that relationships between children and step-parents are often difficult. Like much of evolutionary psychology, Daly and Wilson's work is open to the charge of stating the obvious.

Yet this is also true of much social science that is not informed by what Daly and Wilson like to call "selectional" thinking. Often, the point of doing studies is to firm common sense up with numbers. And Daly and Wilson's numbers are dramatic. Based at McMaster University, in Ontario, they have worked mainly with Canadian statistics. In one study, they found that children under two - the most vulnerable age group - were seventy times more likely to be killed by step-parents than by genetic parents. Though Daly and Wilson tend not to break their figures down by sex, this essentially means stepfathers. There are too few stepmothers of very young children for the statistics to mean much.

All their findings, which draw upon figures from the US and England and Wales as well as Canada, confirm this pattern. For babies and very young children, the risk of being killed by a step-parent is some 50 - 100 times higher than that from a genetic parent. Using US data, Daly and Wilson found the excess risk of non-fatal physical abuse to be around seven times higher, which would look massive were it not set next to the fatal abuse statistics. The conclusion they draw is that steprelationship is "the single most important risk factor for severe child maltreatment yet discovered."

It would certainly be possible to discover this factor without even having heard of Darwin. But Daly and Wilson point out that thinking selectionally led them to an insight which prevailing social science models had obscured. They suggest that an emphasis on notions such as that of roles prevented investigators from focussing on steprelationship in itself. "This really amazes us; I can't quite figure this out to this day," says Daly. "It's got to be more than merely an oversight, it's got to be motivated neglect. No one had asked, 'is there excess risk to children in stepfamilies, and if so, to what degree?' " That risk, he suggests, turns out to be much higher than what social workers might have intuitively suspected.

Worse, they recall, when they began their studies nearly twenty years ago, the limited discussion of stepfamilies in the academic literature had a tendency to blame the children themselves, for provoking hostility by their resentful attitudes.. Even in the late 1980s, a survey listing 89 risk factors for violence against women failed to consider step relationships. Yet surveys conducted at a women's shelter by Lisa Singh, one of Daly and Wilson's students, suggested that women were more vulnerable if not all their children were fathered by their current partners.

Daly and Wilson did not enter this field as social workers. Their previous research subjects were non-human - Daly worked on desert rodents, a research interest he continues to maintain, and Wilson on monkey behavioural endocrinology. Their work on stepfamilies forms part of a wider investigation of killing, discussed in their 1988 book Homicide, initiated as an exploration of "the applicability of contemporary evolutionary theory to the analysis of human motives and perceptions of self-interest". In evolutionary terms, killing one's own descendants looks about as contrary to one's interests as it is possible to get.

There are, however, circumstances in which self-interest appears to dictate infanticide, a widespread phenomenon in human societies. Children who appear unlikely to develop normally, for example, may be killed at birth. In terms of selection, the logic is that the resources needed to raise such a child would be wasted, and under conditions of hardship, it makes more sense to save them for another child.

The killing of unrelated young occurs in other species, such as lions and langur monkeys. Once a male langur has succeeded in the struugle for a sexual monopoly of the females in a troop, he will dispatch the existing infants. But Daly and Wilson do not believe that the tragedies behind their statistics result from similar motives of genetic self-interest. "We don't think we're looking at an adaptation when we look at child killing or assault," Margo Wilson told the Darwin Seminar audience. Rather, these killings are malfunctions of parenthood; mercifully rare, but much commoner in stepfamilies than genetic ones.

The form of the killing appears to be different, too, suggesting different impulses. Stepfathers seem to kill more brutally, suggesting greater aggression and hostility. One of Daly and Wilson's surveys found that 80 per cent of under-fives killed by their stepfathers were beaten to death, a proportion twice as high as that of under-fives killed by their genetic fathers. Daly and Wilson obtained very similar percentages from British statistics. By contrast, genetic parents are more likely to kill themselves after killing their children. That suggests a different mindset; one dominated by depression.    

It might just be that, taken as a whole, the kind of people who become step-parents are more violent than those who don't. But there is evidence that when violent step-parents occupy households which also include children of their own, they spare these and direct their aggression at the stepchildren.

Daly and Wilson emphasise that killing represents "the tip of an iceberg of parental feelings". It is a lot easier to study abuse than kindness. "We only looked at maltreatment because we couldn't get information on any benefits, but we'd expect exactly the same," says Wilson.

They chose to focus on homicide largely because of the reliability of the data: murder is much less likely to go unrecorded than lesser forms of violence. But stepchildren may be at greater risk of death even without family violence. A study of the Ache, a foraging people who live in Paraguay, found that 43 per cent of children raised by a mother and stepfather died before they reached the age of 15. For children raised by both biological parents, the figure was 19 per cent. Under conditions so harsh that one child in five dies, it may take only small differences in the level of care to push the mortality rate among stepchildren towards one in two.

Unsurprisingly, death and serious physical harm in children are associated with hardship in industrial societies as well as forager communities. Severe child abuse cases are rarely detected in middle-class homes. But although poverty is a risk factor, it seems to be independent of the risk associated with stepfamilies. Among the comfortably off, strains in stepfamilies might be expected to produce more moderate effects. It appears, for instance, that middle-class stepchildren in Canada and the US leave home earlier than other children.

By this stage, it may look as though Darwinism is an elaborate way of discovering what not just commonsense has been saying, but Christianity has been teaching all along. In Paraguayan forests and English suburbs alike, marriages that last are best for children. Maybe serial monogamy is, in some sense, against our nature?

At first sight, Daly and Wilson's work seems ripe for plucking by some overexcited conservative moralist. So far from coming to see stepfamilies as unnatural, though, Daly and Wilson are inclined to think that a readiness to become step-parents may be part of our evolutionary heritage. The example of the Ache shows that stepfamilies are hardly unique to the developed West. Daly and Wilson also suggest that in the West, stepfamilies may actually have been more common in the nineteenth and eighteenth centuries, as the result not of divorce, but remarriage following the early death of spouses. Nor, they observe, are humans the only creatures who are prepared to look after juveniles that are not their own offspring. In other species, such as a kind of fish that lives in anemones, this readiness is interpreted as a price that is in the creature's interest to pay, in order to secure an attractive mate.

Similarly, a human's relationship with a stepchild is likely to be based on an understanding of reciprocity. The step-parent contributes to the care of the child in order to secure the relationship with the parent. But his or her fundamental interest is in the parent, not the child. As family therapists recognise, relationships with stepchildren are in some ways more like friendship.

Wilson suggests that an awareness of the risks of step-parenthood might encourage women with children to take greater care in selecting a new partner. "They also might modulate their behaviour to think that it's not their entitlement to have him behave towards this child as if he were the genetic father," she adds; "that they might show more gratitude and reciprocity for the solicitude, and they might be better mediators between the child and the person." It might be natural, but nobody's saying it's easy.