It is generally felt amongst the medical profession and the lay public that cancer is being treated more successfully than in the past. This is certainly true for childhood malignancies and leukaemia but evidence that significantly improved survival is occurring in the common solid tumours is lacking. Since 1963 the University of Liverpool Department of Otolaryngology/Head and Neck Surgery has collected data on all patients with head and neck tumours presenting to the department. The present study investigates patients with histologically proven squamous cell carcinoma of the four main sites: larynx, hypopharynx, oral cavity and oropharynx. From 1963 until the end of 1989, 2738 patients were seen by the department and from 1990 a further 717 patients have been seen. Since 1990 patients have tended to be in better general physical condition but, on the other hand, have tended to have more advanced disease at the primary site. The department has latterly tended to see fewer laryngeal cancers and more cancers of the oropharynx. Significantly fewer patients have presented with neck node metastases. Multiple logistic regression suggests that the most significant difference between the two groups is the great reduction in neck node recurrence rates in the group of patients seen since 1990 (P = 0.0001). The recurrence of tumours at the primary site since 1990 has been 35% compared with 41% before 1990, and recurrence in the neck nodes since 1990 has been 12%, compared with 15% before 1990. These differences are significant (P = 0.0141 and P = 0.0494, respectively). When studying survival in the 1960s, 1970s and 1980s, the 5-year cure rate was 50%, whereas since 1990 the figure has risen to 60% tumour-specific 5-year survival--a significant difference. A similar effect was noted in observed survival. This improvement in cure rate occurred for all four main sites. The results were confirmed by Cox's proportional hazards model where year of treatment was highly significantly associated with improved survival (P = 0.0001). It has been demonstrated that locoregional recurrence has improved since 1990 and this is reflected in improved survival figures. Although there are differences in the parameters of tumours referred before 1990 and since 1990, multivariate analysis suggests that the improvement in neck node recurrence rates may be responsible for this improved survival rate. Multivariate analysis for survival also suggests that the improvement in cure rates is independent of compounding variables and dependent on the year of presentation of the tumour. This improved survival may be related to factors, such as the administration of radical postoperative radiotherapy.