of antisocial personality—so most analyses were performed on counts of antisocial symptoms (e.g., lying, truancy, trouble with the law) in the adoptees. Both an antisocial and an alcoholic biologic background predicted the number of antisocial symptoms, although the significance levels of the prediction varied slightly from publication to publication (Cadoret, 1978; Cadoret and Cain, 1980; Cadoret et al., 1983, 1985, 1986). Later publications stress the independence between genetic liability to antisocial personality and alcohol abuse (Cadoret et al., 1985).

On the environmental side, discontinuous mothering of adoptees, adoptive parent psychopathology, and divorce or separation in the adoptive family predicted antisocial outcomes. Male adoptees appeared more likely than females to have high antisocial behavior counts in the presence of these adverse environments (Cadoret and Cain, 1980). Recently, substance abuse in the adoptees could be predicted from biologic parents' antisocial behavior and alcohol abuse (Cadoret et al., 1986).

Since this review was completed, the preliminary reports in three different samples of twins have also pointed to heritability for antisocial behavior. The first sample was a small series in which twins were ascertained in a psychiatric setting and demonstrated both heritability of symptoms of antisocial personality disorder and important genetic correlations between antisocial behavior and alcohol and substance abuse (Carey, 1993; Miles and Carey, 1993). The second sample was a population-based twin sample of children and adolescents and suggested modest heritability for a classification of symptoms of conduct disorder (Eaves et al., 1993). The final sample consisted of male Vietnam-era armed services veteran twins. Analysis of symptoms of antisocial personality disorder suggested moderate heritability for adult symptoms with a stronger influence of common environment for adolescent symptoms (Lyons et al., 1994).

Only three modern samples permit meaningful analyses of violent offenders—the Danish adoption sample, the Stockholm adoption sample, and the Danish twins. Table 6 presents a summary of their findings. In the Danish adoption study, the only publicly reported data are the proportions of male adoptees who participated in a violent offense as a function of the number of criminal convictions in their biological parents (Mednick et al., 1983, 1984, 1988). Mednick et al. (1983) also cite other analyses (e.g., violent

crimes in parents and in adoptees) that did not uncover a significant relationship, but no data are presented.

From Bohman et al. (1982:Table 5), it is possible to reconstruct contingency tables relating the probability that biological fathers registered for at least one violent offense will have an adopted son registered for crime or alcohol abuse. (The investigators did not publish data on violent offenses in offspring). The association between paternal violence and adoptee's criminal registration in the Stockholm adoption study is slightly negative—7.9 percent of the adopted sons of biological fathers registered for a violent crime are criminal, compared to 10.6 percent of the sons of fathers who were not registered for a violent offense. This negative relationship also appeared for women (Sigvardsson et al., 1982:Table 4).

Paternal violence, however, relates significantly to sons' but not daughters' registration for alcohol abuse. Given a biological father registered for a violent offense, 24.7 percent of the sons exhibited alcohol abuse; if the biological father did not have a

violent offense on record, 16.6 percent of the sons were registered for alcohol problems (χ² = 4.145, p < .05, tetrachoric r = .145).

Finally, Cloninger and Gottesman (1987) analyzed the male MZ and same-sexed male DZ Danish twin series by crimes against persons, a category previously defined by Christiansen to include crimes of violence and sexual offenses. In contrast to the adoption studies, the twin data demonstrate a strong familial effect for these crimes. Concordances for the 24 MZ and 39 DZ probands were, respectively, 41.7 and 20.5 percent, compared to respective base rates of 3.3 and 2.8 percent among the two types of twins. The tetrachoric correlations for both zygosities are significantly greater than 0.0: r_MZ = .77 ± .11 and r_DZ = .52 ± .10. Although they suggest a genetic effect, a likelihood ratio test for heritability is not significant (χ²₁ = 1.775, .20 < p < .10).

Major limitations of all three studies are the reliance on official records and insufficient detail in any of the studies to assess the magnitude of the correlation between recorded crimes and violent acts as defined by the panel. For example, would a significant proportion of crimes that involved an assault be registered under a Scandinavian equivalent of disturbing the peace? Inaccuracies of classification would make it difficult to uncover heritable effects and could be used to explain the adoption results. If violence had the same approximate biological father adopted offspring correlation as property crime (about .15 judging from Baker et al., 1989) and the base rates for violence range from about 6 to 10 percent among biological fathers (Mednick et al., 1983) and 2 to 5 percent among their adopted sons, then the Danish sample has between .55 and .82 power to detect an effect at the .05 level. On the other hand, if inaccuracy in classification lowers the correlation to .10, power is diminished to .25–.45. The fact that twin concordance is significant ensures that the record classification is not totally unreliable but at the same time does not permit actual quantification of reliability.

Together the data do not suggest a strong role for heredity in violence. On the one hand, the positive correlation between violence in biological parents and alcohol abuse in adopted sons and the trend of the twin correlations suggest a genetic effect. On the other hand, the failure in both adoption studies to detect a significant relationship between violent offending and other indices of crime in separated relatives is evidence that any putative genetic factor is weak. Whatever the case for genetics, the strong correlations in MZ and DZ twins suggest that family factors shared by siblings would be a profitable avenue for future research.