For histological subtyping of anal squamous carcinomas the WHO advocates a six-way subdivision, but it has been suspected that the six types cannot be reliably discriminated in practice. We conducted a blinded study involving slides from 103 consecutive cases, each slide being examined by three experts (from Denmark, Australia and UK) on two occasions at least 8 months apart. Agreement on subtypes was low: 72% between rounds within pathologist, 61% between pathologists. Even for the commonest, and most stably diagnosed, type, viz. large-cell keratinising squamous carcinoma, the intra- and interpathologist frequencies of confirmation were only 81% and 71%, respectively. The pathologist marked the picture as typical and his subtype diagnosis as certain 41% of times: even then confirmation frequencies were only 88% and 74%, respectively. Calculations, including kappa analyses, suggest that 26% of the typing variation was noise. The WHO scheme must be even more unreliable in everyday practice. We finally mention a recently demonstrated link between human papilloma virus (HPV) and certain types of anal cancer, which may well provide an additional argument for revising existing subtyping schemes.