violence, crime, and other psychopathology. Early research compared the prevalence of sex chromosome aneuploidy among selected samples (e.g., prisoners, psychiatric patients) with the prevalence among newborn screens or other controls. The initial reviews of 47,XYY and 48,XXYY syndromes by Owen (1972) and Hook (1973) stressed that prospective research on well-defined populations was necessary for accurate assessment of any relationship between XYY and violence.

Several prospective studies have now followed aneuploid children into adolescence. Results on 47,XYY suggest behavioral development well within the normal range, but with minor deficits in intelligence, other cognitive skills, and perhaps, in emotional and social skills (Bender et al., 1987). Study of a Danish birth cohort (Witkin et al., 1976) found a high prevalence of criminal registration among XYY individuals (5 of 12) that is not statistically different from the prevalence among 47,XXY (3 of 16) but is greater than the base rate among normal XY males of the same height (9% of slightly more than 4,000 men). Criminal histories of the five XYY individuals were not characterized by violence and aggression. Convictions were for minor offenses adjudicated by mild penalties, prompting the investigators to suggest that the relationship was likely due to nonspecific factors such as lowered intelligence. Personal follow-up revealed that the 12 XYY individuals had statistically significant but clinically minor differences from controls in sexuality, aggression, and testosterone levels (Schiavi et al., 1984, 1988; Theilgaard, 1984).

The prospective results dispel the myth of the XYY as a "hyperaggressive, supermasculine sociopath" and, in its place, portray a group of individuals within the normal range but with an array of relatively nonspecific behavioral differences in attention and cognition, motoric skills, and personality. For example, the sexuality of XYY individuals is characterized more by insecurity and difficulty in developing and maintaining satisfying relationships with women than by stereotyped hypermasculinity (Schiavi et al., 1988). It is possible that nonspecific behavioral problems may increase risk among these individuals for later criminal offenses.

Given that genes do not code directly for crime and violence, it might be reasonable to suspect that genetic diathesis is mediated through personality traits and cognitive styles. Kinship correlations for intelligence have been summarized by Bouchard and

McGue (1981), and the general pattern suggests important contributions from both genes and family environment. A survey of the genetic literature on personality traits is too vast to report here. Hence, this review is limited to two major domains of personality. First, scales purporting to measure aggression are reported, with the name of the scale dictating acceptance of a study for review. In some studies, the exact items bear strong content resemblance to the concept of violence used by the Panel on the Understanding and Control of Violent Behavior (e.g., Tellegen et al., 1988); in other cases, the content appears related more to a broader, almost psychoanalytic, notion of intrapunitiveness (e.g., Partanen et al., 1966); and in yet other studies, item content was not fully specified (e.g., Rushton et al., 1986). The second area for review is scales specifically constructed to predict juvenile delinquency. These scales include the Psychopathic-deviate (Pd) scale of the Minnesota Multiphasic Personality Inventory (MMPI; Hathaway and McKinley, 1940); the Socialization (So) scale of the California Psychological Inventory (CPI; Gough, 1969); and the aggression scale of the Missouri Children's Picture Series (MCPS).

Table 2 summarizes the results of this review. The studies are broadly classified into twin and adoption strategies. The twin results are consistent with the overall twin literature on personality: on the average, identical twins correlate higher than fraternal twins. A notable exception is reported by Plomin et al. (1981), the only study using blind ratings of aggression in a test situation (ratings of child's aggression against a Bobo doll). All other studies used self-report inventories or parental ratings. Unfortunately, sample sizes in Plomin et al. (1981) are too small to detect whether this difference is attributable to method of assessment, to aggression in childhood versus adulthood, or to sampling error from a trait with modest heritability.

In the Minnesota series of twins raised apart (Gottesman et al., 1984; Tellegen et al., 1988), the correlations are as great as those for adult twins raised in the same household. Although standard errors for these correlations are large, they suggest that the great similarity of twins raised together is not due exclusively to such processes as imitation or reciprocal interaction (Carey, 1986) that might invalidate the twin design.

There is little relevant adoption data. Different scales were administered to the Texas Adoption Project sample (Loehlin et al., 1985, 1987) in adolescence (the CPI in Loehlin et al., 1985) and in early adulthood (the MMPI in Loehlin et al., 1987). The patterning here is very similar to that of other adoption studies