were predicted by the number of convictions among biological fathers and antisocial types of psychiatric disorder in biological mothers.

In addition to his study of unwanted pregnancies, Bohman (1978) identified 2,324 individuals born in Stockholm, Sweden, between 1930 and 1949 who were given up by birth parents for nonfamilial adoption. These adoptees and their birth parents were tracked through official Swedish records for alcohol abuse (having a fine imposed for intemperance, supervision by a local temperance board, or treatment for alcohol abuse) and criminality (defined as a conviction with more than 60 "day fines," a fine prorated to the defendant's income). At the time of the record search, adoptees' ages ranged from their early twenties to early forties.

Like the biological parents in the Danish sample, biological mothers and fathers both had high rates of registration, but one striking feature distinguishes these two Scandinavian samples—none of the adoptive parents in Sweden appeared on the criminal register (Bohman et al., 1982), whereas the Danish adoptive parents had rates of criminal registration only slightly below population base rates (Mednick et al., 1983). Thus, the Swedish adoptees may be regarded as a special sample selected for higher-than-average genetic liability but lower-than-average family environment liability to crime. What effect did this peculiar selection have? Apparently, the effects cancel each other. The base rate for criminal registration among male adoptees was 12 percent compared to an 11 percent population risk (Bohman et al., 1982). This fact alone argues against misguided views of genetic determinism of behavior that is not amenable to environmental intervention.

The initial results were strikingly negative for any genetic effects on crime: 12.5 percent of the male adoptees with a criminal biological father themselves had a criminal record, compared to 12 percent of the male adoptees whose biological fathers did not have criminal records. Respective prevalences for the adoptees with and without criminal mothers were 12.6 and 12.4 percent. Base rates for criminal conviction among females were too low to permit meaningful analysis.

More detailed analyses on 76 percent of this adoptive sample were reported by Bohman et al. (1982), Cloninger et al. (1982), and Sigvardsson et al. (1982). (The reduction in sample size was due to deletion of adoptees with incomplete information, late placement, or intrafamilial adoption.) For male adoptees, there was a

nonsignificant correlation between adoptee conviction and biological parent conviction—13.2 percent of 258 male adoptees with either a convicted father or a convicted mother were themselves convicted, compared to 10.4 percent of 604 adoptees whose biological parents had not been convicted (χ²₁ = 1.37, p = NS). Analogous figures for female adoptees were not reported.

Suspecting heterogeneity for criminality, Bohman et al. (1982) conducted a series of discriminant analyses aimed at distinguishing subgroups of male adoptees. The final classification suggested that alcohol abuse moderated the genetic relationship to criminality. Criminal male adoptees who were also registered for alcohol abuse tended to have committed a larger number of offenses, were more likely to have been convicted of a violent offense, received longer jail sentences, but were less often registered for property crimes than male adoptees registered only for criminality. Moreover, biological parental variables also distinguished the male adoptee subgroups. Most important here are the findings that violent paternal offenses predicted offspring alcohol abuse better than they predicted offspring criminality and that adoptees registered for criminality but not for alcohol abuse tended to have fathers registered for property crimes and fraud. On the environmental side, male adoptees with criminality only were distinguished by late placement in adoptive homes, the number of placements, and a longer duration spent with the biological mother relative to other adoptee groups.

Others (Walters and White, 1989) have criticized this study for not reporting and controlling for the number of statistical tests, but the original investigators did partly replicate the discriminant function results on female adoptees (Sigvardsson et al., 1982). Of greater concern is what the results mean.

The inability to uncover statistically significant findings on criminality per se is a clear failure to replicate the results in neighboring Denmark. Perhaps this is due to differences in registration practices between the two countries. Was there a tendency in Sweden to report instances of joint alcohol abuse and illegal activity to the temperance board, and not to the police, with perhaps the reverse tendency in Denmark? A second possibility is that the selection of adoptive households in Sweden was so extreme that it overcame potential genetic liability for crime or violence. The pattern might occur if the relationship among genotype, family environment, and crime were nonlinear or if there were strong interaction between genotype and family environment. The pattern of cross-fostering risk in the Stockholm