Home :: Research Focus Boxes :: Research Strategies :: Matched–Subjects Design

Matched-Subjects Design: Sexual Orientation of Children Raised in Lesbian Families

How much, if at all, do parents influence the sexual orientation of their children? How likely is it that a child’s sexual orientation will be affected by observing the gender roles of parents? Are sexual preferences learned through exposure to parents who model various behaviors? Are they acquired through a process of identifying with the same-sex parent? Or do sexual preferences reflect, in large measure, a roll of the genetic dice?

Research with gay men with either monozygotic or dizygotic twin brothers (Bailey & Pillard, 1991), and with lesbian women with twin sisters (Bailey, Pillard, Neale, & Agyei, 1993), finds that monozygotic twins are significantly more likely to also be homosexual themselves than are dizygotic twins. This finding indicates a genetic link to homosexuality in that monozygotic twins have identical genetic makeups, developing as they do from the same fertilized ovum, whereas dizygotic twins develop from two separate ova and are no more similar genetically than any other two siblings. One’s genetic makeup, then, can be argued to significantly influence one’s sexual orientation.

Social-cognitive theory, on the other hand, argues that modeling of gender roles by the same-sex parent, along with the parents’ differential reinforcement of appropriately masculine or feminine behaviors, are important determinants of gender development. Accordingly, family environment, as provided by the sexual orientation of parents, should be an important influence on the sexual orientation of children.

Previous research on the development of sexual orientation in children has been limited to heterosexual families. Critical comparisons of the sexual preferences of adult children raised by homosexual versus heterosexual parents have been lacking. Susan Golombok and Fiona Tasker (1996) corrected this deficiency by investigating the sexual preferences of children raised in lesbian families and by comparing these with the preferences of children raised in heterosexual families.

Golombok and Tasker contacted 27 lesbian mothers and 27 heterosexual mothers when their children were approximately 9 years old. The children were contacted again 14 years later, when they were approximately 23 years old, to determine the influence of being raised by a lesbian mother. Simple? Yes, as long as one controls for other variables that might relate to sexual orientation, such as the presence or absence of a father in the home, mother’s age, and social class. Golombok and Tasker made sure that children in each of the two types of families were raised in homes in which the fathers were absent, the homes differing only in the sexual orientation of their mothers. Thus the lesbian mothers were matched with control mothers who differed only in sexual orientation. (As one might expect, most of these mothers had at least one relationship [heterosexual] during the time the children were living at home, just as most of the lesbian mothers had at least one lesbian relationship.)

To match subjects along some variable, one first needs to rank the subjects in each sample according to the matching variable—say, social class—and then draw pairs of subjects from the two samples that are from the same or approximately the same income bracket. Using this procedure, one can be sure that the two groups will be equivalent regarding the matching variable. If social class is related to sexual orientation, it will be equated for the two groups. Any differences between groups in sexual orientation cannot be due to social class.

Matching carries an additional advantage: It reduces the amount of unexplained variability in the groups. This variability is termed random error. By reducing random error, one can more easily see the effects of the variable of interest. Another way of describing this advantage is to say that matched designs are more sensitive than those in which each subject is randomly assigned. The sensitivity of a design refers to its ability to detect a difference due to the treatment variable if such a difference exists.

Matching sounds like such a good idea that one has to wonder why investigators don’t use this method all the time. Yet, like other procedures, it has its disadvantages. The most serious drawback is a statistical one concerning the degrees of freedom used when determining the significance of the tests that evaluate the research outcome.

In designs that do not match subjects, the degrees of freedom reflect the number of subjects; in matched-subjects designs, they reflect the number of pairs. Matching cuts the degrees of freedom in half. This means that one must obtain a larger difference for it to reach statistical significance. The irony to this disadvantage is that matching is most advantageous when one is using few subjects, because it increases the sensitivity of the design. But this is the very condition under which one can least afford to lose degrees of freedom. Before one matches, one needs to be sure that the matching variable is highly correlated with the measure one is using. Only in this way will it effectively reduce unexplained variability and pay for the loss in degrees of freedom.

Matched designs are slightly more time consuming to conduct than are those involving simple random assignment of subjects. One must first administer the matching variable, then rank subjects before they can be assigned to conditions. Extra expense may also be involved. A more serious disadvantage than either of these is the threat to external validity that occurs when subjects who can’t be matched must be discarded. Any loss of subjects can affect the representativeness of the sample and the ability to generalize to the population from which it was drawn.

What does Golombok and Tasker’s matched design tell us about the influence of lesbian mothers on the sexual orientation of their children? Although children from lesbian families more frequently explored same-sex relationships, by far most of the children of lesbian families identified themselves as heterosexual.



J. M. Bailey & R. C. Pillard. (1991). A genetic study of male sexual orientation. Archives of General Psychiatry, 48: 1089–1096.

J. M. Bailey, R. C. Pillard, M. C. Neale, & Y. Agyei. (1993). Heritable factors influence sexual orientation in women. Archives of General Psychiatry, 50: 217–223.

S. Golombok & F. Tasker. (1996). Do parents influence the sexual orientation of their children? Findings from a longitudinal study of lesbian families. Developmental Psychology, 32: 3–11.