This study tests the hypothesis that the reported relationship between life event stress and physical illness is primarily a function of criterion and other content contamination in the stress measure. In particular, conventional life event measures include events related to physical health, which overlap with the criterion; events related to neuroticism, which influences the criterion; and events that are vague or subjective and could be affected by individual differences in psychological distress, response sets, and retrospective bias. In this study 386 adult males and females completed standard measures of life events and physical illness. Illness was significantly related to event subscales containing, respectively, health-related events, neuroticism-related events, and subjective events, but not to an "uncontaminated" event subscale. These results support the hypothesis of contamination and suggest that alternative approaches to the conceptualization and measurement of stress may need to be developed.