In 1998, Storey and co-workers suggested that individuals homozygous for arginine (Arg) at codon 72 of the p53 gene are about seven times more susceptible to human papillomavirus (HPV)-related carcinogenesis than heterozygotes. Since then, several studies from Northern Europe, Japan and the USA have failed to demonstrate a similar correlation. By contrast, a study in Brazil as well as one recent study in Italian and Swedish populations showed strong positive associations. We examined the frequency of p53 codon 72 polymorphism in samples from both invasive and intra-epithelial cervical neoplasias (CIN), and compared them with samples from healthy controls. All 88 samples came from women with a Greek ethnic background. Tissue specimens were collected from archival material with histologically diagnosed low-grade CIN (LGCIN), high-grade CIN (HGCIN) or cervical cancer (CxCa). As a control, we used cellular material newly collected by cytobrush from the cervices of 30 healthy women with normal cytological and colposcopical examinations. p53 Arg homozygosity (Arg/Arg) alone was associated with four-, six- or eight-fold increased risks for LGCIN, HGCIN or invasive cancer, respectively. The frequency of the p53Arg/Arg genotype and of the proline (Pro) allele showed significant linear trends according to the degree of severity of the lesion (P = 0.0007 and P = 0.0009, respectively). Exclusion of the ten HPV16/18-negative cases did not substantially alter the Arg/Arg frequency among the groups nor the significant linear trend. Our results confirm the initial findings of Storey and co-workers, as well as the data of the Brazilian and the recent European study, but do not accord with those of the other aforementioned studies. Variations in ethnic background, laboratory performance, verification of the HPV status, definition of controls, and sample size are the most plausible explanations for this controversy. In all our samples, the distribution of the p53 alleles fits the Hardy-Weinberg equilibrium and the 0.48 frequency of the Pro allele in our controls accords well with the percentages previously reported for different ethnic groups as characteristic of the assumed north-south cline. Some authors assert that the discrepancy in the results could not be attributed to differences in the methods; however, the Brazilian study emphasized the effect of inter-laboratory variation in detecting the association between p53 polymorphism and cervical cancer. Regarding the control group, our samples were only from women with a cytologically and colposcopically benign cervical epithelium. We think that simply choosing 'normal volunteers' for collecting control DNA blood samples without knowing the status of their cervical epithelium is indeed a possible source of bias. Finally, it is very unlikely that loss of heterozygosity at the p53 locus could be a factor interfering with the allelotype distribution. Our present small study results, which suggest a biologically relevant association, provide strong evidence that homozygous arginine at codon 72 of p53 may confer a higher susceptibility to HPV-associated intra-epithelial and invasive cervical neoplasia.