may account for these discrepancies. First, the types of food additives used in these studies have varied substantially. Second, a wide ranges of doses have been used. It has been argued that the dose of food additives used in some experiments was too low to produce adverse behavioral effects. However, it should be noted that the results of studies using larger doses have been both positive and negative. Finally, age may alter sensitivity to food additives. In general, younger children have been found to be more sensitive to food additives than older children (Harley et al., 1978; Conners, 1980; Weiss et al., 1980).

The preceding data allow several inferences to be drawn about the effects of food additives on hyperactive behavior. First, Feingold's claims and those from other open trials have been overstated. At best, only a small percentage of hyperactive children may be adversely affected by food additives. Second, younger children may be more sensitive to food additives than older children. Third, there may be a dose-response curve for food additives, just as there is for any toxic substance, but this has yet to be demonstrated.

In conclusion, the data on food additives and behavior are such to preclude any major legislative or administrative action to remove food additives or severely limit their use. Further studies of those few children who appear to respond negatively to food additives seem warranted. Additionally, research with experimental animals examining whether food additives have any biological activity in the central nervous system is recommended.

Over the past 10 years, the results of several types of experiments have suggested that an inverse relationship may exist between blood cholesterol concentration and violent behavior (Virkkunen, 1983b; Virkkunen and Penttinen, 1984; Kaplan and Manuck, 1990; Muldoon et al., 1990). For example, Virkkunen and colleagues (Virkkunen, 1983b; Virkkunen and Penttinen, 1984) measured fasting serum cholesterol levels in male homicidal offenders and found that those with antisocial personality disorder or intermittent explosive disorder with habitually violent tendencies had lower cholesterol levels than other offenders. This difference was particularly pronounced in men under the age of 30. It was hypothesized that the lower cholesterol levels in the violent offenders may be a consequence of enhanced insulin secretion.

Although the proposed relation between cholesterol levels and violent behavior is provocative, it suffers from the same problems as the proposed association between hypoglycemia and violence. No information on dietary history is provided in any of the studies by Virkkunen and colleagues. Additionally, the role of alcohol in influencing both cholesterol level and violent behavior is not addressed. Moreover, other differences between the violent offenders and other offenders (e.g. activity levels) are never explored.

To further investigate the relation between cholesterol level and aggressive behavior, Muldoon and colleagues (1990) compared the causes of mortality for subjects in intervention groups and control groups in six large primary prevention trials for reducing cholesterol levels. Cholesterol reduction in the intervention groups was accomplished in two studies by nutritional manipulations aimed at reducing dietary cholesterol and saturated fat intake and in the remaining four studies by pharmacological treatment. Causes of mortality were divided into three categories: coronary heart disease, cancer, and causes not related to illness, which included deaths due to accident, suicide, or homicide. In all studies, the treatments led to significant reductions in cholesterol levels. Compared to controls, the average cholesterol concentration of participants in the intervention groups was reduced by approximately 10 percent. This reduction in cholesterol was not associated with a significant decline in total mortality. However, cholesterol reduction was associated with a lower mortality rate from coronary heart disease and, in some studies, with a slightly higher mortality rate from cancer. With regard to mortality not related to illness, the chance of dying from suicide or violence was approximately twice as great in the intervention groups than in the control groups. The association between lower cholesterol levels and increased mortality from causes other than illness was found regardless of whether lipid lowering was based on dietary or pharmacological treatment.

Although the results of a number of studies have suggested an inverse relation between cholesterol levels and deaths due to accidents or violence, this association has not been universally observed. For example, Pekkanen et al. (1989) found little evidence of an inverse association between serum cholesterol values and increased mortality due to accidents and deaths in their 25 year follow-up of 1,580 Finnish men.

Additional epidemiological investigations and experimental research are clearly required to assess the relationship between blood cholesterol level and violent behavior.