A key question in research linking psychosocial factors with disease and health is whether psychosocial factors add to the effects of known risk factors or are interactive with them. Eysenck would have us believe that certain psychosocial factors, notably personality and stress, act synergistically with certain risk factorsthat is, each by itself is relatively benign, but their effects multiply to produce high levels of disease. Researchers and clinicians concerned with the relationship between psychosocial factors and health would be pleased to believe that the interaction hypothesis (in general) and the synergistic hypothesis (in particular) are true. Personality and would qualify being variables in behavioral medicine, and psychologists would be assured roles in the funding of research and the practice of health care. The risk-factor interaction hypothesis is a special case of the biobehavioral interaction hypothesis, a fundamental hypothesis in behavioral medicine (Weiss & Schwartz, 1983). The hypothesis is that biological and psychosocial variables are not merely additive classes of factors that contribute independently to health and illness, but they can interact with each other (e.g., synergistically) and therefore act interdependently. The biobehavioral interaction hypothesis is, in turn, a special case of what may be called the generic interaction hypothesis, which is a fundamental hypothesis in systems theory in general (Bertalanffy, 1968) and living systems theory in particular (Miller, 1978). The concept of the whole being more than the sum of its parts (e.g., interactive rather than additive) is essential to the notion of a system. It is from this notion that the idea of an emergent property comes forth. All systems evidence novel properties that emerge from the interaction of specific subsystems, systems, and suprasystems. It has been proposed that concepts and methods from systems theory provide a powerful integrative tool for researchers and clinicians in behavioral medicine (Schwartz, 1977, 1979, 1980, 1982, 1984, 1987, 1990). Eysenck proposes that psychosocial variables are as important and equal to smoking, heredity, cholesterol level, blood pressure, and other physical variables risk factors for cancer and coronary heart disease (CHD). Risk of cancer and CHD can be greatly diminished through behavioral therapies, and psychological influences on physical diseases are much greater than heretofore suspected. What is the basis for drawing such strong conclusions? I put aside serious questions regarding Eysenck's conceptions of personality types and stress, and their associated measurement techniques (addressed in other commentaries), and assume for the moment that the frameworks and tools have some merit. The first set of data bearing on the interaction hypothesis is presented in Eysenck's Table 4 and is graphed here Figure 1. These data, based on a population of 1,341 patients from Yugoslavia and 1,036 patients from Heidelberg, provide strong evidence for an interactive relationship between smoking and Type 1 characteristics. The second set of data bearing on the interaction hypothesis is presented in Eysenck's Table 9 and is graphed there his Figure 1. These data, based on a matched-proband population of 512 patients from Heidelberg, provide some evidence for an interactive relationship between numbers of cigarettes smoked and (defined having a higher score for Types 1 + 2 + 5 than for Types 3 + 4 + 5). It is only at 41 to 60 cigarettes per day that the presence of is more than additive (i.e., interactive). It is unclear if levels of are indeed comparable across numbers of cigarettes smoked. The third set of data bearing on the interaction hypothesis is presented in Eysenck's Table 10 and is graphed here Figure 2. These data, based on the same matched-proband population of 512 patients from Heidelberg, provide weak evidence for an interactive relationship between numbers of close relatives who died or were suffering from lung cancer and stress measured by Eysenck. Only at five relatives is the presence of more than additive. At six relatives, the presence of is associated with double the percentage deaths from lung cancer. The levels of across numbers of relatives are not provided. One might expect that with increases in numbers of relatives dying from lung cancer might come increases in levels of stress. The fourth set of data bearing the interaction hypothesis is presented in Eysenck's Table 11 and is graphed here Figure 3. These data, based on a population of 871 patients from Heidelberg, appear to provide evidence for an interactive relationship between numbers of risk factors (heredity, cigarettes, bronchitis, and stress-here defined probands of Types 1 or 2) and death from lung cancer. Whereas single risk factors and pairs of risk factors have relatively little association with deaths from lung cancer, combinations of three factors and especially the combination of the four factors has a strong association with deaths from lung cancer. However,