Head and neck cancers are a significant and worsening health problem in the UK. Minimising diagnostic delay could result in diagnosis at an earlier stage of the disease, which can usually provide a better prognosis. The delay from the patient's experience of the symptoms to the ultimate diagnosis consists of the patient delay plus the General Practitioner interval. This study aimed to establish the period of delay between recognition of the initial tumour symptoms and the formal diagnosis, in order to understand the relative contribution of the patient and physician. 133 males and 55 females, mean age 65 years (range 24-94 years) were interviewed by a research nurse, using a piloted semi-structured questionnaire. Tumour size at diagnosis was classified according to T1 (20%), T2 (29%), T3 (27%), T4 (22%). Cancer of the lip and oral cavity (CLOC) had occurred in 48 (26%) patients and 138 patients had other head and neck cancers. The first symptom included change in voice (26%), pain (27%), lump or growth (12%) and dysphagia (8%). Other symptoms were mentioned by 37 (26%) patients and included infection, sore throat, ulcers and abscess. The median number of weeks of delay from the onset of symptoms to the patient's initial decision to seek professional advice was 6 weeks for all patients, 6 weeks for CLOC patients but 5 weeks for other head and neck patients. For 30 patients it was 9 weeks to 6 months before they sought help. The median delay from the onset of symptoms to the patients being seen by a hospital consultant was 8 weeks for all patients but 12 weeks for patients with CLOC. Whilst 34 (25%) did not find their initial symptoms at all worrying, the remainder had some concerns, but 6 (4%) patients only suspected cancer. No significant association was found between the nature of the first symptoms and the urgency with which patients interpreted their symptoms, the diagnostic delay with the patients' gender, age and social class. However, patients with CLOC were reported to have longer delays than other patients. It is concluded that the contribution of the patient to the total length of delay was found to be significant, especially with CLOC patients for whom the trend towards a greater delay is a cause of concern. The absence of any association in this study may relate to the particular habits of this high-risk group, perhaps associated with other aspects of their social behaviour. This finding warrants further investigation.