Sequences are presented for the signal peptides of prelysozymes from 6 species of birds and compared to the known sequence for chicken prelysozyme c. The sequencing was done with synthetic oligonucleotides as primers and oviduct mRNA as the template, obviating the need to clone DNA from these species. Ring-necked pheasant prelysozyme c differs from all other prelysozymes c and pre-alpha-lactalbumins examined by being cleaved in vivo between amino acid residues 17 and 18 instead of between residues 18 and 19. The feature unique to the signal peptide of pheasant prelysozyme c is proline at position 17. Besides showing that proline is acceptable as the carboxyl-terminal amino acid of the signal peptide, our finding implies that it cannot occur as the penultimate amino acid in the signal peptide. This supports the view that unless a polypeptide has the proper secondary structure, signal peptidase will not cleave it, and that this secondary structure is a beta-turn. Another outcome of this comparative study is an estimate that the mean rate of sequence evolution in the prelysozyme signal peptide is 1%/two million years of divergence, similar to that calculated for the insulin signal peptide. Because this rate is a third of the silent substitution rate, it is likely that one out of every three amino acid substitutions is compatible with signal peptide function.