The tumour suppressor p53 is a transcription factor that binds DNA in the vicinity of the genes it controls. The affinity of p53 for specific binding sites relative to other DNA sequences is an inherent driving force for specificity, all other things being equal. We measured the binding affinities of systematically mutated consensus p53 DNA-binding sequences using automated fluorescence anisotropy titrations. Based on measurements of the effects of every possible single base-pair substitution of a consensus sequence, we defined the DNA sequence with the highest affinity for full-length p53 and quantified the effects of deviation from it on the strength of protein-DNA interaction. The contributions of individual nucleotides were to a first approximation independent and additive. But, in some cases we observed significant deviations from additivity. Based on affinity data, we constructed a binding predictor that mirrored the existing p53 consensus sequence definition. We used it to search for high-affinity binding sites in the genome and to predict the effects of single-nucleotide polymorphisms in these sites. Although there was some correlation between the K(d) and biological function, the spread of the K(d)s by itself was not sufficient to explain the activation of different pathways by changes in p53 concentration alone.