Disagreement persists about the superiority of Reagan and Ng's method over that of Broders' for the histologic grading of squamous carcinoma of the cervix. Uncertainty about the predictive value and reproducibility of any of the grading methods prompted a comparison of factors previously suggested as indicating the biologic behavior for cervical squamous carcinoma. One hundred ninety-five women, who were enrolled in a Gynecologic Oncology Group treatment protocol of Stage IB squamous carcinoma of the cervix and underwent radical hysterectomy with pelvic and paraaortic node sampling, formed the study population. The tumors were graded first by participating institutional pathologists, with submitted slides subjected to an independent review by two pathologists (R.J.Z. and S.W.). The histologic parameters examined included the presence and amount of keratinization, nuclear pleomorphism, mitotic rate, gestalt grading, pattern of invasion at the stromal interface, and inflammatory cell infiltrate. The depth of invasion and presence or absence of vascular invasion also were assessed. The probability of pelvic lymph node metastasis and the progression-free interval were determined for each parameter. Surprisingly, none of the grading methods was effective in predicting nodal spread or progression-free interval. However, an increasing depth of invasion strongly correlated with nodal spread and a diminished progression-free interval (P less than 0.0001). Vascular invasion was less effective in these predictions (0.05 less than P less than 0.10). Both measurements were reasonable reproducible. It was concluded that histologic grading of surgically treated cervical carcinoma is not useful but that the depth of invasion and vascular invasion are important predictors of behavior that should be reported routinely.