The deduced amino acid sequence at the hemagglutinin (HA) cleavage site of 76 avian influenza (AI) viruses, subtypes H5 and H7, was determined by reverse transcription-polymerase chain reaction and cycle sequencing techniques to assess pathogenicity. Eighteen of the 76 viruses were isolated in 1993 and 1994 from various sources in the United States. In addition, 34 H5 (4 highly pathogenic [HP] and 30 non-highly pathogenic [non-HP]) and 24 H7 (3 HP and 21 non-HP) repository viruses, isolated between 1927 and 1992, were sequenced and the sequences compared to those in recent isolates. All repository HP H5 and H7 viruses studied had multiple basic amino acids adjacent to the HA cleavage site and most had basic amino acids in excess of the proposed minimum motif B-X-B-R (B = basic amino acids arginine or lysine, X = nonbasic amino acid, R = arginine) that has been associated with high pathogenicity. Of the non-HP viruses studied, 35 of 38 for H5 and 30 of 31 for H7 conformed to the motif B-X-X-R and B-X-R, respectively. Two non-HP H5 viruses had the motif X-X-X-R at the cleavage site and a third had the motif B-X-X-K (K = basic amino acid lysine). One non-HP H7 (A/Pekin robin/CA/30412-5/94) had four basic amino acids (K-R-R-R) adjacent to the cleavage site. Although the Pekin robin isolate did not produce disease in chickens under the conditions of the study it did have the amino acid sequence compatible with that in HP AI viruses and, therefore, is considered potentially HP. This is the first account of an H7 virus that is non-HP in chickens but meets the molecular criterion to be classified as HP.