Many eukaryotic transcriptional activator proteins, including the Xenopus 5S RNA gene activator protein TFIIIA and the HeLa cell protein Sp1, have an approximately 30 amino acid repeating motif which binds to short, specific DNA sequences. Over 150 of these sequences are now known. Based on the observed distribution of amino acid residues, a series of constraints and predictions can be proposed for the structure of the motif. A compatible three-dimensional structural model has been developed by a combination of interactive model building and refinement by molecular dynamics. The model structure consists of a two-stranded beta-hairpin stabilizing a C-terminal alpha-helix by both zinc ligands and hydrophobic interactions. Four of the residue positions on the helix N-terminus and exposed face are predicted to provide base specific ligands. Further implications of the model for DNA binding are discussed.