To provide an epidemiologic test of the “spectrum of disease” concept in cervical neoplasia, a sample of invasive cancer, in situ cancer, severe dysplasia, and mild-to-moderate dysplasia cases and matched controls was drawn from the women examined through a cervical cancer screening program for medically indigent women. All women with the appropriate histologic diagnoses comprised the case groups and, from the pool of women with normal smears, 4 controls per case were selected, so that 4 controls matched each case on age, race, and time period of examination. Among the variables for which cases and controls were compared, an abnormal cervix identified at the screening examination was more frequent among all categories of cases than among controls. Postcoital and postmenopausal bleeding differentiated invasive cancer cases from controls, and both categories of cancer (in situ and invasive) differed from controls on the basis of lack of postnatal care and need for referral for medical care, as identified at the screening examination. Both categories of dysplasia patients differed from their controls in terms of the increased frequency with which the patients were receiving public welfare. From these associations, the demonstrated lack of association with other variables, and the lack of a gradient in the strength of association for individual variables from the least to the most severe category of neoplasia, the spectrum of disease hypothesis can neither be confirmed nor refuted by the current study.