their followers with their oratory and their own brand of proselytism. Two of the most prominent leaders of this movement, Sylvester Graham (remembered best for the development of the graham craker) and John Harvey Kellogg (recognized for the introduction of precooked breakfast cereals), lectured widely throughout this country promoting the use of natural foods and decrying the ingestion of meat, which they believed would lead to the deterioration of mental functioning and the arousal of animal passions (Whorton, 1982). Kellogg further concluded that the breakdown products of meat acted as dangerous toxins that, when absorbed from the colon, produced a variety of symptoms including depression, fatigue, headache, and aggression. Kellogg wrote that "the secret of nine-tenths of all chronic ills from which civilized human beings suffer" including "national inefficiency" and "moral and social maladies" could be traced to the meat eater's sluggish bowels (Kellogg, 1919:87). In keeping with his puritanical background, Kellogg also warned his followers that spicy or rich foods would lead to moral deterioration and acts of violence (Kellogg, 1882:244–245).

The foregoing historical information should not be seen as simply humorous background material. Ideas about food and behavior continue to be prevalent. The last decade has witnessed an explosion of interest in the field of nutrition and behavior. The current obsession with health and fitness, as well as the desire to use diet as a panacea, has led to a myriad of dietary "self-help" books. Unfortunately, the consumer's desire for simple answers to complex questions has often led to misinterpretation or even misrepresentation of scientific data. Correlational data have been interpreted as signifying cause and effect relations. With the public spotlight focused so strongly on the area of nutrition and behavior, it is crucial that research in this area be based on proper methodology and careful interpretation of data. This is particularly true for studies examining the relationship between diet and antisocial behavior because policy decisions may be made on the basis of this research.

One of the more difficult problems in research on diet and behavior is how to separate nutritional from nonnutritional factors. Because food is so intimately involved with other aspects of our daily lives, it contains much more than its obvious nutritional

value. Food is an intrinsic part of social functions, religious observations, and cultural rituals. Because food is a "loaded" variable, both experimenters and subjects may harbor biases about expected research outcomes. To minimize the confounding effects of these biases, double-blind procedures in which neither the experimenter nor the subjects know what treatment is given must be used.

In general, research on nutrition and behavior would benefit from the methodological controls used in psychopharmacology (Dews, 1986). Two variables that are important in drug studies are dose and length of treatment. Because consumption of a small amount (dose) of a dietary component may produce behavioral effects that differ from consumption of a larger amount, several doses of a dietary component should be tested whenever feasible. By using different doses, researchers can determine if there is a systematic relationship between the dietary variable and behavior. The lack of a systematic effect should be taken as a danger sign either that the apparent effect is spurious or that the variability is greater than expected.

The duration of dietary treatment is also critical. Although short-term (acute) studies permit evaluation of the immediate effects of a dietary treatment, they cannot provide information about long-term (chronic) exposure. Because the behavioral effects of dietary components (e.g., food additives) may only appear with extended exposure, both acute and chronic studies should be used to assess the nutrition-behavior interaction.

Diurnal variations in subjects' responses to nutrients should also be considered in diet-behavior studies. For example, it was recently observed that a snack (candy bar or yogurt) significantly improved subjects' ability to pay attention to relevant stimuli when it was eaten in the late afternoon, but not when it was consumed in the late morning (Kanarek and Swinney, 1990).

Prior nutritional status also has the potential of influencing the results of acute experiments. The types and amounts of foods previously consumed can affect the metabolism of a test nutrient. Standardizing dietary intake prior to evaluating the behavioral consequences of a test nutrient can eliminate this source of variation.

Another challenge in planning nutrition and behavior experiments is choosing appropriate subjects. Differences in nutritional history, socioeconomic background, and other environmental factors create subject heterogeneity that poses a threat to the internal validity of the research. Internal validity concerns the ability