PURPOSE: To analyze the degree of tumor extension in T2 and T3 laryngeal cancer, independently of vocal cord fixation, and to introduce a new CT grading system for use in preoperative T-stage assessment. MATERIALS AND METHODS: Retrospective analysis of degree of tumor extension was performed in 36 patients with laryngeal cancer (T2 glottic, 4 ; T2 supraglottic, 12 ; T3 glottic, 12 ; and T3 supraglottic, 8). T-stage was determined according to clinical and pathologic findings, and based on the TNM classification of AJCC (1992). The degree of tumor extension seen on CT was determined by the number of involved anatomic subsites and compared with T-stage and lymph node metastasis. On the basis of statistical analysis(Fisher's exact test) of those results, we suggest new CT grading system for laryngeal cancer. RESULTS: Fifteen of 20 supraglottic cancer patients showed six or less involved anatomic subsites ; twelve of these 15 (80%) were at stage T2 and three (20%) were T3. Four of the 15 showed lymph node metastasis. The five patients whose number of involved anatomic subsites was seven or more were all T3 ; four (80%) of these showed lymph node metastasis. The difference in the number of involved anatomic subsites in T2 and T3 tumor was statistically significantly different (p<0.05) ; the incidence of LN metastasis was more prevalent in the group with seven or more involved subsites. In glottic cancer, however, no statistically significant difference was observed between T2 and T3 tumors in the number of involved subsites. CONCLUSION: In supraglottic cancer, T1 and T4 stages are determined by degree of tumor extension, irrespective of vocal cord mobility ; T2 and T3 stages are suggested by the number of involved anatomic subsites, without reference to vocal cord mobility. A new CT grading system may thus be made, based on degree of tumor extension. In glottic cancer, however, laryngoscopy or dynamic study with spiral CT are needed for T-stage assessment.