We have previously reported that in four transcription factor families the DNA-recognition rules can be described as (i) chemical rules, which list possible pairings between the 20 amino acid residues and the four DNA bases, and (ii) stereochemical rules, which describe the base and amino acid positions in contact. We have incorporated these rules into a computer program and examined the nature of the rules. Here we conclude that the DNA recognition rules are simple, logical, and consistent. The rules are specific enough to predict DNA-binding characteristics from a protein sequence.