1. There is a trend in most studies, albeit not always a statistically significant one, consistent with the hypothesis of a genetic effect on adult and perhaps adolescent antisocial behavior. The consistency of the literature is not to be confused with generalizability of the results to contemporary violence in the United States. The latter issue is discussed in point 10.

2. The genetics of antisocial behavior do not fit a simple model. Here, a central comparison involves the estimates of heritability from the Danish twin and adoption studies, given that sample sizes are large and that cross-national variations in definitions of crime and its registration are minimized. By using tetrachoric correlations, the twin data suggest a heritability of around 55 percent. Correlations from the Danish adoption data (Baker et al., 1989) suggest heritabilities about one-half this amount. On a simple additive genetic model, both estimates should be similar. The joint effects of marital assortment, temporal trends over time, nonadditive genetic variance, special twin effects, etc., must be considered in studying the genetics of antisocial behavior.

3. There is reason to question the application of simple genetic models to the traditional twin design in studying the genetics of adult antisocial behavior. Both the Danish and the Norwegian twin series provide evidence of imitation, collusion, or reciprocal interaction between twins, especially MZ twins, for criminal participation. Rowe (1985) also reported that adolescent twins often engage in delinquent acts together. Although collusion might invalidate the assumptions of the traditional twin design, the application of statistical models that include imitation effects (Carey, 1986; Eaves, 1976) or direct measurement of potential interactive effects (e.g., the analyses performed by Rowe, 1985) might illuminate important sources of sibling similarity that cannot be isolated without the genetic information provided by twins. Along these lines, data from single-offspring families and data expressed as a function of sibship constellations must complement the twin data (see Carey, 1986; Eaves, 1976).

4. The adoption studies have consistently reported a correlation between adoptee antisocial behavior and some variable intervening between birth and final placement in the adoptive home. Some preplacement variable has predicted adoptee antisocial behavior in the Crowe, Cadoret, Danish, and two Swedish adoption studies. The relevant construct, however, eludes identification because the same variable (e.g., number of placements

versus disruptive mothering) has not been replicated across studies or within genders of a single study (e.g., compare Bohman et al., 1982, with Sigvardsson et al., 1982).

5. Evidence for a genetic effect specific to offenses involving physical aggression is weak. Part of the problem is that the relatively low base rate for violent offenses such as murder, assault, and rape requires samples much larger than those already gathered to give meaningful and replicable results. What is clear is that genetic liability toward participation in criminal activity is not due exclusively to liability toward physical aggression; otherwise, one would find little heritability of liability for crime per se but rather a tendency for violent crime to aggregate in twins and biological relatives of violent adoptees.

6. The available evidence is inconsistent with a major polymorphism on the Y chromosome associated with antisocial behavior. a major polymorphic gene on the Y chromosome could theoretically explain some of the large gender differences in criminal participation and also some of the heritable individual differences within males. Although the Y chromosome appears to contribute to aggressive behavior in the mouse (Maxson et al., 1989), a similar polymorphism in humans would predict greater father-son than father-daughter or mother-son resemblance. The tetrachoric correlations for the Danish adoption study (see Baker et al., 1989) do not show this pattern.

7. The selection of adoptive families may make it difficult, in adoption study designs, to generalize about the effects of family environment that are relevant for criminal participation of offspring. Registered criminality among adoptive parents was non-existent in the Swedish adoption study; antisocial behavior was rare among the adoptive parents in Cadoret's study; and a five-year period free of criminal registration was a requirement for adoptive families in Denmark. These findings are consistent with some selection against prospective adoptive parents with a criminal record and raise the possibility that adoptive homes are non-representative with respect to family environmental factors that contribute to offspring antisocial liability. Under these circumstances, it would be unwise to attribute the lack of adoptive parent-offspring correlation for criminality to an absence of vertical environmental transmission. The study of unselected, intact nuclear families in concert with the biological and adoptive families would provide important information about vertical transmission, but such control families have not been utilized in most adoption research.

8. There is evidence for a genetic association between antisocial behavior and alcohol abuse and possibly other substance abuse. The Danish adoption and twin studies did not analyze substance abuse, and the Norwegian twin study, although it gathered ered data on alcohol use, did not report analyses of its association with criminality in twins. Crowe reported no excess of alcohol abuse among the adopted offspring of female felons. The Swedish study reports a genetic association between biological paternal violence and adoptee alcohol abuse, and the investigators suggested genetic heterogeneity—the joint occurrence of criminality and alcohol abuse may have a different genetic liability than that of criminality alone. The association in Cadoret's series varies slightly with the type of data analysis, but points in directions similar to those of the Stockholm study. There is also a strong suspicion that alcohol abuse is itself genetically heterogeneous, with one meaningful subgroup related to antisocial personality (Cloninger, 1987).

9. There is a strong tendency, especially in the area of juvenile antisocial behavior, to implicate some correlate of home environment as an etiological component in antisocial behavior. Analyses of four different samples—Cadoret et al. (1975), Cadoret (1978), Crowe (1972) as analyzed by Cadoret et al. (1983), and Bohman (1971)—reveal significant findings in this area. The base rate for criminality in the adult Stockholm adoptees is also congruent with the hypothesis that adoption neutralized genetic liability. Although the expanded Danish sample did not reveal a significant correlation with adoptive parent criminality, urban environment predicted adoptee crime independent of biological background (Gabreilli and Mednick, 1984).

10. It is difficult to integrate the genetic literature on antisocial behavior into contemporary criminological research on violence. There are two major reasons for this. The first is extrapolation over space and time; the second is extrapolation over measures of violence. A considerable amount of current research in the United States targets high-crime areas, usually urban, and uses survey instruments. There is no sample in the genetic literature that can readily extrapolate to an urban U.S. population studied with survey instruments.

With regard to extrapolation over time and over cultures, it must be recalled that heritability is a function of the amount of environmental variability for a trait. As environmental variance decreases, heritability increases, and as environmental variance increases, heritability decreases. It is theoretically possible that

increased rates of violence associated with drug abuse, for example, might reduce the magnitude of a genetic effect on antisocial behavior or even change the nature of that effect (e.g., loci that contribute to the reinforcing effects of cocaine may become more important). The larger samples in the genetic literature are Scandinavian. One could argue that their relative cultural and socioeconomic homogeneity would provide a ripe medium for maximizing genetic effects. In addition, the Scandinavian samples involve cohorts that went through the major portion of risk for initial antisocial behavior before the upsurge of drug use in the 1960s and 1970s. (All the Danish twins went through their risk period before World War II.)

Another limitation of the large Scandinavian studies is the exclusive reliance on official records for defining phenotypes such as criminality and violence. Although ethics may preclude personal contact with retrospectively identified adoptive families, field research that compares classification via official records with classification using contemporary survey instruments would increase the interpretability of the Scandinavian results. If such research has already been conducted, the results have not been integrated into the primary reports on the adoptees and twins.

The results of the smaller U.S. studies, especially Crowe's, are very consistent with the Scandinavian research. In two of the studies, Cadoret's and Rowe's, the phenotypes include counts of adolescent antisocial behaviors that include many peccadillos (e.g., lying, truancy) when compared to an adult court conviction (used in the Danish study). Taken at face value, the Rowe and Cadoret results suggest heritability of ''mild" antisocial behavior. If there were cross-national generalization of these results, one would expect the Scandinavian material to show a similar pattern. Arrests have apparently been recorded in the Danish study (see Mednick et al., 1988) but have not been subjected to genetic analysis. Also, two of the U.S. samples were conducted in Iowa, and the third noted few respondents from inner-city areas (Rowe, 1983).

Can we extrapolate these results to major U.S. metropolitan areas? In the primary genetic literature, there is no attempt on the part of the authors to provide information congruent or discrepant with such extrapolation. Hence, there is no positive evidence to permit this generalization, but at the same time, there is no direct negative evidence against the induction. So, given the consistency of the genetic literature, it is perhaps best to conclude that it would be unwise to overlook the possibility of a genetic contribution to individual differences in antisocial behavior in contemporary criminological research.