One hundred and thirty-five patients with squamous carcinoma of the vulva were treated at UCLA and City of Hope Medical Centers between 1957 and 1985. Sixty-two cases were stage I, 48 stage II, 18 stage III, and 7 stage IV. Twenty-one patients developed a local vulvar recurrence after primary radical resection. Ninety-one patients had a surgical tumor-free margin greater than or equal to 8 mm on tissue section and none had a local vulvar recurrence. Forty-four patients had a margin less than 8 mm; 21 had a local recurrence and 23 did not (P less than 0.0001). Of the 23 patients with a margin less than 8 mm who did not recur locally, 14 remained free of disease, and 9 had either advanced disease, declining health, or short follow-up. Depth of invasion is associated with local recurrence, with a 9.1-mm reference value correctly predicting outcome in 81.5% of cases. Increasing tumor thickness is associated with local recurrence, with a 10-mm reference value predictive of 90% non-recurrence and 33% recurrences. A pushing border pattern is less likely to recur than an infiltrative growth pattern. Lymph-vascular space invasion has a combined predictive accuracy of 81.5%. Increasing keratin and greater than 10 mitoses per 10 high-power fields correlate with local recurrence. Neither clinical tumor size nor coexisting benign vulvar pathology correlates with local recurrence. Fourteen of twenty-one patients with vulvar recurrence died of metastatic disease, four died of intercurrent disease, and three were alive at 32, 68, and 157 months, with 16 recurring in less than 1 year. Surgical margin is the most powerful predictor of local vulvar recurrence. Combining factors in a stepwise logistical regression does not significantly improve this predictive value. Accounting for specimen preparation and fixation, a 1-cm tumor-free surgical margin on the vulva results in a high rate of local control, whereas a margin less than 8 mm is associated with a 50% chance of recurrence.