Two hundred cases of early squamous cell carcinoma of the cervix were studied, 109 with early stromal invasion (ESI) (FIGO stage Ia1) and 91 microcarcinomas (MCs) (stage Ia2). Nine percent of invasive foci arose from original (native) squamous epithelium, outside the transformation zone. What may have been "skip lesions" (see text) were seen in only two instances. Of the cases of ESI, 78% invaded to < or = 0.5 mm. Of the MCs, 65% invaded to < or = 3 mm, and > 50% measured < or = 50 mm3 in volume. Capillary-like space involvement (CLSI) was suspected in 32 cases (29%) of MC on hematoxylin and eosin (H&E) staining but could not be confirmed by Ulex europeaus agglutinin I lectin immunoperoxidase staining in 10. Forty-four MCs displayed a spray-like growth pattern, 34 a confluent pattern, and 12 both; one had a condylomatous appearance. Forty-eight MCs were grade 3, 27 were grade 2, and 16 were grade 1 in differentiation. Stromal response was assessed as absent in nine cases, weak in 53, moderate in 58, and intense in 80. The extent of associated cervical intraepithelial neoplasia (CIN) was widespread in all but nine cases. There were seven recurrences, including one definite and one possible tumor-related death. A single case of residual disease was observed. The new FIGO classification of preclinical stage Ia carcinoma as that diagnosed only by microscopy is sound. The division into stages Ia1 and Ia2 is reproducible and may be used for comparison between various institutions; the notion, however, that the length of MC should form part of its definition is not supported by these data. None of the variables (depth of invasion, length, area, volume, growth pattern, grade, stromal response, and CLSI) had any bearing on prognosis. Although stromal invasion can certainly be seen in small punch biopsies, a definitive diagnosis can be made only in conization (or hysterectomy) specimens. It is absolutely essential that conization specimens be thoroughly sampled, not only to make the correct diagnosis but also to be certain about the margins. Early stromal invasion behaves in the same manner as CIN and may be treated by conization alone provided the cone has been adequately sampled and the margins are free. Much the same applied to MC, although there were not enough patients whose tumors were associated with CLSI to be absolutely certain of the prognostic significance of the latter.