The development of an H2N2 vaccine is a priority in pandemic preparedness planning. We previously showed that a single dose of a cold-adapted (ca) H2N2 live attenuated influenza vaccine (LAIV) based on the influenza A/Ann Arbor/6/60 (AA ca) virus was immunogenic and efficacious in mice and ferrets. However, in a Phase I clinical trial, viral replication was restricted and immunogenicity was poor. In this study, we compared the replication of four H2N2 LAIV candidate viruses, AA ca, A/Tecumseh/3/67 (TEC67 ca), and two variants of A/Japan/305/57 (JAP57 ca) in three non-human primate (NHP) species: African green monkeys (AGM), cynomolgus macaques (CM) and rhesus macaques (RM). One JAP57 ca virus had glutamine and glycine at HA amino acid positions 226 and 228 (Q-G) that binds to α2-3 linked sialic acids, and one had leucine and serine that binds to α2-3 and α2-6 linked residues (L-S). The replication of all ca viruses was restricted, with low titers detected in the upper respiratory tract of all NHP species, however replication was detected in significantly more CMs than AGMs. The JAP57 ca Q-G and TEC67 ca viruses replicated in a significantly higher percentage of NHPs than the AA ca virus, with the TEC67 ca virus recovered from the greatest percentage of animals. Altering the receptor specificity of the JAP57 ca virus from α2-3 to both α2-3 and α2-6 linked sialic acid residues did not significantly increase the number of animals infected or the titer to which the virus replicated. Taken together, our data show that in NHPs the AA ca virus more closely reflects the human experience than mice or ferret studies. We suggest that CMs and RMs may be the preferred species for evaluating H2N2 LAIV viruses, and the TEC67 ca virus may be the most promising H2N2 LAIV candidate for further evaluation.