The human transcriptional enhancer factor (TEF) family includes TEF-1, TEF-3, TEF-4, and TEF-5. The TEFs share a highly conserved 68-amino acid TEA/ATTS DNA-binding domain, which binds to SV40 GTIIC (GGAATG), SphI (AGTATG), SphII (AGCATG), and muscle-specific M-CAT (GGTATG) enhansons. We determined the optimal DNA-binding consensus sequence for TEF-1. Using a purified GST-TEF-1 fusion protein and a random pool of synthetic oligonucleotides, 31 independent clones were obtained after six rounds of binding site selection. DNA sequences analysis revealed that 16 clones contained direct repeats with a 3-bp spacer (DR3), and 15 clones contained a single binding site. The predominate consensus half-site was GGAATG (67%), and the other elements were of the form G_AGA_T/CATG. The TEF-1 bound to the DR3 as a dimer in a cooperative manner. Cooperative binding was dependent on the spacing and orientation of the half-sites and was inhibited by deoxycholate treatment, providing evidence that protein–protein interactions were involved. The data suggest that TEF dimerization is important for its ability to modulate gene transcription.