The aim of this study is to present and discuss the results of initial radiotherapy with salvage surgery and to compare them with those from centres where primary surgery is the prevailing principle. The series comprised 1005 consecutive patients treated during the period 1965-1998. Salvage surgery was performed if patients had residual tumour or developed recurrence. Disease-specific survival (DSS) and crude survival (CS) after 5 years, among 643 patients with glottic carcinomas treated with curative radiotherapy was 88.6% (SE = 1.3) and 65.3% (SE = 2.0), respectively. Among T1 glottic carcinomas the locoregional control was 88%, i.e. 88% of patients were cured after radiotherapy alone, and the DSS was 99% both evaluated after 5 years, i.e. the salvage surgery added approximately 11% to the survival of T1 glottic patients. Only 4% (12/312) of T1 glottic patients had laryngectomies. Locoregional control among T2 glottic cases was 67% and the DSS 88%, but, 18% (41/233) of patients lost their larynx. The corresponding results among T3 glottic cases were 30% and 59%, i.e. the organ preservation was close to 50%. Among patients with supraglottic carcinomas, the two estimates were 440 and 63%, respectively. Compared with our results, recent results published in the literature after initial laser surgery of T1 glottic carcinomas indicate that there are only minor differences in DSS and organ preservation, but it is generally agreed, but not proven, that voice quality after radiotherapy is better. T2 glottic carcinomas treated by initial supracricoid partial laryngectomy in a selected series have yielded very high DSS rates with better organ preservation than was observed in the present series. As to T3 glottic carcinomas, initial surgery does not produce better survival rates than those produced in the present series but our organ preservation is higher. The treatment of patients with supraglottic carcinoma has benefited from optimization of radiotherapy during recent decades. The role of initial laser surgery is as yet undecided.