Protein sequences evolve through random mutagenesis with selection for optimal fitness. Cooperative folding into a stable tertiary structure is one aspect of fitness, but evolutionary selection ultimately operates on function, not on structure. In an accompanying paper (Socolich et al., 2005), we proposed a model for the evolutionary constraint on a small protein interaction module (the WW domain) through application of the SCA, a statistical analysis of multiple sequence alignments. Construction of artificial protein sequences directed only by the SCA showed that the information extracted by this analysis is sufficient to engineer the WW fold at atomic resolution. Here, we demonstrate that these artificial WW sequences function like their natural counterparts, showing class- specific recognition of proline-containing target peptides. Consistent with SCA predictions, a distributed network of residues mediates functional specificity in WW domains. The ability to recapitulate natural-like function in designed sequences shows that a relatively small quantity of sequence information is sufficient to specify the global energetics of amino acid interactions.