Stress researchers, in attempting to circumvent methodological problems in correlation studies, have resorted to analytical techniques that increase the likelihood of a Type II error. For instance, studies suggesting that stress plays no role in irritable bowel syndrome are at variance with a large body of clinical experience and may be in error. Several measurement issues are discussed, but the most significant concerns the way investigators deal with individual differences in chronic levels of stress and symptoms. J. Suls and colleagues, in this issue, recommend correcting for these individual differences before estimating the relationship between stress and symptoms, but this is illogical because most stressors are chronic--they are related to the social and economic circumstances in which people live. A more important confound, not considered by most investigators, is the contribution of somatization to stress assessment and symptom reports.