Solution structures of two Src homology 3 (SH3) domain-ligand complexes have been determined by nuclear magnetic resonance. Each complex consists of the SH3 domain and a nine-residue proline-rich peptide selected from a large library of ligands prepared by combinatorial synthesis. The bound ligands adopt a left-handed polyproline type II (PPII) helix, although the amino to carboxyl directionalities of their helices are opposite. The peptide orientation is determined by a salt bridge formed by the terminal arginine residues of the ligands and the conserved aspartate-99 of the SH3 domain. Residues at positions 3, 4, 6, and 7 of both peptides also intercalate into the ligand-binding site; however, the respective proline and nonproline residues show exchanged binding positions in the two complexes. These structural results led to a model for the interactions of SH3 domains with proline-rich peptides that can be used to predict critical residues in complexes of unknown structure. The model was used to identify correctly both the binding orientation and the contact and noncontact residues of a peptide derived from the nucleotide exchange factor Sos in association with the amino-terminal SH3 domain of the adaptor protein Grb2.