(e.g., Scarr et al., 1981)—zero-order correlations with adoptive parents and small positive correlations with biological parents. This is consistent with the twin data in suggesting heritability, although estimates of the genetic effect are lower in the adoption than in the twin studies. The fact that adoptive relatives bear little resemblance to one another suggests that processes such as imitation and common family environment have weak effects on these psychometric predictors of delinquency. Parker's (1989) recent analysis of data from the Colorado Adoption Project (Plomin et al., 1981) gives a very different picture. Based on maternal ratings of aggression, both adoptive and biological siblings show strong, roughly equal, resemblance. These data agree with the Plomin et al. (1981) results in suggesting important common environment effects for childhood aggression, but do not agree with the small twin studies of Owen and Sines (1970) and Ghodsian-Carpey and Baker (1987).

Together, the personality data imply a genetic contribution to individual differences for important correlates of violence. This inference is stronger for older adolescents and adults than for children. A glaring lack in this literature (as well as in the genetic literature on criminal offenders) is the absence of data to permit multivariate genetic analysis of personality traits such as aggression and criminal offending.

Since this review was completed, several important studies of parental ratings of childhood aggression and delinquency in child and adolescent twins have been completed and initial results suggest important heritability. Gottesman and Goldsmith (in press) should be consulted for a review.

Here, research on actual crime or antisocial behavior during adolescence is reviewed. Early twin studies of juvenile delinquency, summarized in Table 3, gave strong evidence of important common environment effects and weak evidence for heritability. Many of these early studies did not select samples or define phenotypes with the rigor required by modern research standards (see Slater and Cowie, 1971; Christiansen, 1977, for reviews). More recently, Rowe (1983, 1985, 1986) analyzed the number of self-reported antisocial behaviors from junior and senior high school twins. The twin correlations (given previously in Table 2) demonstrate significant heritability and, agreeing with the concordances

in Table 3, implicate common environment, albeit without reaching statistical significance.

Adoption studies support the importance of family environment in early antisocial behavior. Cadoret et al. (1983) reported a significant main effect for an adverse adoptive home and some form of gene-environment interaction in three different adoptee samples. Bohman (1971) and colleagues (Bohman and Sigvardsson, 1985; Bohman et al., 1982) prospectively studied Swedish children from unwanted pregnancies. At age 15, those who remained with their own biological parents or who were placed as foster, but unadopted, children had almost twice the rate of antisocial-like behavior problems (truancy, running away, misuse of alcohol and drugs, repeated thefts) as their classmate controls. Children who were formally adopted, however, showed slightly lower rates than their controls and much lower rates than both the foster children and those remaining with their parents, despite a high frequency of criminality and alcohol abuse among their birth parents. Bohman suggested that any adverse genetic liability was neutralized by the benefits provided by secure adoption.

Together, the early twin studies, the nonsignificant trend in Rowe's study, the Cadoret analyses, and the Bohman results provide strong evidence for a family environment effect on juvenile antisocial behavior. Both the Rowe and the later Cadoret studies suggest that genetics cannot be ignored during this period. Perhaps the most important research question for the future is the investigation of genetic and family environmental contributions to adolescent antisocial behavior that persists into adulthood and the detection of reasons why this behavior ceases in many adolescents.