Summary Microscopic malignancy grading using the 8-factor system proposed by Jakobsson et al. (1973), the 4-factor system set up by Glanz and Eichhorn (1985), and DNA cytofluorometry were applied to thirteen Ti and thirty-seven T2 squamous cell carcinomas of the oral cavity, 9 with and 41 without metastases. There was a significant correlation between the presence of lymph node metastases (Ni) and the malignancy scores (P<0.05) and tumour DNA ploidy (P<0.01, chi-square). The total number of patients with initial and late lymph node metastases correlated significantly with polyploid nuclei (P<0.05) and with malignancy scores (P<0.001), which also correlated with regional recurrences (P<0.01, chi-square). No remaining tumour after preoperative radiotherapy indicated less risk for local recurrence than if tumour persisted (P<0.01, chisquare).The cumulative survival (Kaplan-Meier) was worse for patients with nodal involvement (Nl) than for those without (NO) (P<0.01), and for patients with poorly differentiated tumours compared with moderately well differentiated (P<0.05) and to well differentiated (P<0.001). The prognosis was worse for patients with high malignancy scores than for those with low (P<0.001). DNA diploid tumours had a better prognosis than DNA non-diploid, but the difference was not significant.Despite the various modes of treatment available, patients with oral cavity carcinoma pose serious therapeutic problems which are reflected in the poor survival rates (Frazell et al., 1962). Factors influencing the prognosis are therefore sought after. The site and size of the primary tumour and the presence of metastases have been used as prognostic indicators (Lee et al., 1972;Krause et al., 1973;Fletcher, 1979). The presence of cervical lymph node metastases seems to be the most important predictor associated with approximately 50 per cent reduction of the 2-year determinate survival rate (Hibbert et al., 1983; Teichgraeber et al., 1973;Willen et al., 1975; Lund et al., 1975) have cavity carcinoma provides useful prognostic information but has certain limitations. The single method treatment of stage I cancer carries a much poorer prognosis than was previously thought, and the incidence of microscopic cervical metastases is high (Lee et al., 1972;Krause et al., 1973;Teichgraeber et al., 1984).The malignancy grading system based on 4 different morphological characteristics for the tumour cell population, and 4 characteristics for the tumour-host relationship, initially used in the analysis of laryngeal cancer (Jakobsson et al., 1973), has also been applied in oral cavity carcinomas.Studies on palatal, gingival, and lingual carcinomas Willen et al., 1975; Lund et al., 1975) have disclosed statistically significant differences in survival for patients with high and low malignancy scores. The somewhat modified malignancy grading used in lingual carcinomas by Holm et al. (1982) showed a correlation between the malignancy score on the one hand and T classification or the presence of lymph node metastases at th...