Tissue microarrays allow the simultaneous analysis of many tumours using small-diameter cores sampled from larger blocks of tissue, but may be limited by tumour heterogeneity. This study considers the validation of tissue microarray for the study of four molecules of interest as prognostic factors in head and neck squamous carcinoma, including a consideration of methods for assessing immunocytochemical scoring of microarrays. Tissue microarray blocks were constructed from 100 cases of head and neck squamous carcinoma, taking four cores from different areas of each tumour. Immunocytochemical labelling was performed for cutaneous fatty acid binding protein, involucrin, vascular endothelial growth factor and Ki-67. The extent and intensity of scoring was determined for each core and the degree of agreement determined for results from the assessment of two, three or four cores for each carcinoma. In a subset of 30 representative cases, the labelling in the tissue microarrays was compared with that in whole-tissue sections of the same carcinomas. An adequate sample of carcinoma was achieved in more than 90% of the 400 cores; unsuccessful results were attributed to uneven core alignment or to poor targeting of the tumour tissue in the donor blocks. The degree of agreement in the assessment of extent and intensity of labelling was moderate to good (weighted kappa, range 0.479-0.902) between whole-tissue sections and microarray sections depending on the antigen and the scoring system. Tissue microarray is a reliable tool to demonstrate cellular and molecular alterations in head and neck squamous carcinomas. We recommend using the mean results from four cores for biological studies, with analysis of categorical data based on quartile groups. Concordance with whole-tissue section data is reassuring, but data from microarrays need to be validated against clinical outcomes.