Studies of allelic imbalance and suppression of tumourigenicity have consistently suggested that the short arm of chromosome three (3p) harbours tumour suppressor genes (TSGs) whose inactivation leads to the development of various types of neoplasia including head and neck squamous cell carcinoma (HNSCC). Previously, we defined a critical minimal region of 120kb at 3p21.3 that contains overlapping homozygous deletions in lung and breast tumour lines and isolated eight genes from the minimal region. Mutation analysis in a large panel of lung and breast cancers revealed only rare mutations, but the majority of lung tumour lines showed loss of expression for one of the eight genes (RASSF1A) due to hypermethylation of a CpG island in the promoter region of RASSF1A. We found RASSF1A to be methylated in the majority of lung tumours, but to a lesser extent in breast and ovarian tumours. In order to define the role of 3p TSGs, in particular RASSF1A in HNSCC, we (a) analysed 43 primary HNSCC for allelic loss in regions proposed to contain 3p TSGs (3p25-26, 3p24, 3p21-22, 3p14 and 3p12), (b) analysed 24 HNSCC for evidence of RASSF1A methylation and (c) undertook mutation analysis of RASSF1A in HNSCC. We found that 81% of HNSCC showed allele loss at one or more 3p markers, 66% demonstrated loss for 3p21.3 markers and 56% showed allelic losses at 3p12 loci. Thus, 3p loss is common in HNSCC and extensive 3p loss occurs even in early stage tumours. RASSF1A promoter region hypermethylation was found in 17% (4/24) of the sporadic HNSCC, but RASSF1A mutations were not identified. Furthermore, we found RASSF1A methylation to be significantly higher in poorly differentiated then in moderate to well differentiated HNSCC (P=0.0048). Three of the four tumours showing RASSF1A methylation also underwent 3p21.3 allelic loss, hence RASSF1A behaves as a classical TSG (two hits, methylation and loss). One tumour with RASSF1A methylation had retention of markers at 3p providing further evidence of specific inactivation of RASSF1A as a critical step in some HNSCC. Although the frequency of 3p21.3 allele loss was substantially higher than that of RASSF1A methylation this does not necessarily suggest that other genes from 3p21.3 are also implicated in HNSCC, as 3p21.3 LOH was invariably found with LOH at other 3p loci. Thus, the presence of 3p21.3 allele loss without RASSF1A methylation might reflect a propensity for 3p21.3 loss to occur as a secondary consequence of large 3p deletions targeted at other 3p TSG regions. Furthermore, in the presence of homozygous inactivation of other 3p TSGs, RASSF1A haploinsufficiency might be sufficient to promote tumourigenesis in many HNSCC.