Tumour deposits in the head and neck region were treated with hyperthermia using 915 MHz external microwave applicators and radiation therapy between 1986 and 1990. The mean (+/- SE) radiation dose was 47 +/- 2 Gy (range 21-77 Gy). All but four patients had failed previous therapy. Mean tumour volume was 40 +/- 10 cm3 (range 0.3-276 cm3). Hyperthermia was administered biweekly in 80% of the patients in 6.0 +/- 0.4 sessions (range 1-10); thermometry involved 3.6 +/- 0.4 catheters (range 1-9) and 5.7 +/- 0.4 sensors (range 1-12) per tumour. Of the 50 lesions evaluable for response, 29 had a complete response (58%), and 20 had a partial response (40%). Lesions were stratified by depth. In tumours considered potentially heatable (i.e. depth < or = 3 cm and lateral dimensions at least 2 cm less than boundary of applicator), the complete response rate was 81% (26/32, 47 +/- 2 Gy, 15 +/- 3 cm3); whereas for patients with tumours deeper than 3 cm, the complete response rate was 17% (3/18, 48 +/- 3 Gy, 110 +/- 21 cm3), p = 0.0001. Among lesions < or = 3 cm depth that exhibited a complete response, six recurred (24%, 5.8 +/- 1.8 months) while 20 lesions were recurrence free at last follow-up of 11.9 +/- 1.2 months). The overall survival of patients with lesions < or = 3 cm depth was 11.5 +/- 1.3 months (range 2.4-32.3 months) while for patients with lesions > 3 cm depth survival was 6.7 +/- 0.9 months (range 2.1-18.6 months), p = 0.01. In superficial lesions with depth < or = 3 cm, multivariate logistic regression analysis indicated that the model best correlating with complete response included radiation dose (p = 0.08) and tumour volume (p = 0.08, model p = 0.004). Multivariate proportional hazard analysis indicated that the model best correlating with duration of local control included tumour depth (p = 0.03) and previous radiation therapy (p = 0.08, model p = 0.006). Twenty-two fields were treated without any skin reactions (39%), 23 evidenced erythema (40%) and eight thermal blistering (14%). Ulceration occurred in 11 treatment fields but in all but one of these cases the ulceration may have been due to tumour breakdown as there was direct invasion of the skin by tumour prior to the initiation of treatment. The maximal skin temperature was the best predictor of morbidity although the correlation was not statistically significant (p = 0.19).