Traditional sampling methods for the diagnosis of oral candidosis in head and neck cancer patients, i.e. saliva collection or tongue scrapings, are often impossible to perform. The aim was to determine the optimal sampling method. Eighteen oral cancer patients and five control subjects were sampled semi-quantitatively from the labial sulcus, dorsum of the tongue, dental plaque and saliva for cultivation of yeasts. The patients were examined prior to all cancer treatment (n=5), or 2–4 weeks (n=5) or 8–12 weeks (n=8) post-operatively. The incidence of Candida was found to increase from 40 % at the control and pre-operative level up to 73 % 8–12 weeks post-operatively. Candida albicans was found to be the only species until 4 weeks post-operatively. Thereafter, the incidence of species other than C. albicans was 38 %. The most sensitive sampling site was found to be the vestibular sulcus, from which all culture-positive cases could be confirmed. Tongue surface scraping was found to be more sensitive than saliva collection in detecting Candida. All sampling methods and sites were equally sensitive in detecting the different Candida species. Dental plaque was found to have the highest density of Candida colonization, and was thus found to be the most significant source of Candida infection, which emphasizes the role of dental care in these patients.