BACKGROUND: Little is known about the diagnosis of head and neck carcinoma in primary care. We sought to estimate the general prevalence of symptoms reported by patients with head and neck carcinomas and to determine the association between detection patterns of head and neck cancer cases in primary care and survival. METHODS: In a cross-sectional survey, we used a questionnaire to estimate the general prevalence of symptoms associated with head and neck cancer from a sample of 5646 primary care visits in 25 randomly selected health centres over 4 weeks throughout Finland. A population-based retrospective cohort study involved the 221 patients resident in one primary health care district (population about 700,000) in whom head and neck carcinoma was diagnosed between Jan. 1, 1986, and Dec. 31, 1996. Data on the initial primary care visit, clinical characteristics and survival were obtained from patient charts. RESULTS: Of 5646 visits to a primary care practitioner, 11% (617) were made because of the same symptoms as those initially reported by patients later found to have head and neck cancer. According to the cohort data, the detection rate of these carcinomas in primary care was 1 per 63,000 visits. At the initial visit of 221 patients later found to have cancer, 56% (123) received referrals, 24% (53) follow-up appointments and 20% (45) neither ("overlooked"). At 3 years, the risk of death was significantly higher among patients whose disease was overlooked (adjusted hazard ratio [HR] 1.89, 95% confidence interval [CI] 1.03-3.45). The excess risk associated with being overlooked, however, was confined to subjects with tongue or glottic tumours (HR 4.25, 95% CI 1.59- 11.4) (number needed to harm 3.0, 95% CI 1.9-6.7). INTERPRETATION: Despite the rarity of patients with head and neck carcinoma in primary care, patients with symptoms of these diseases and especially with symptoms of tongue and glottic carcinomas should be initially referred for further care or followed up.