Japanese macaques (Macaca fuscata) and control species (vervet, pigtailed macaque, bonnet macaque) were trained for food to respond to one class of recorded fuscata vocalizations and to not respond to a second class. A measure of neural lateralization was obtained by presenting the stimuli randomly to the right or the left ear, and comparing performance in the two ears (ear advantage method). Vocalizations were from Steven Green''s field tapes. In experiment I, the two classes were Green''s ''smooth early high coos'' (SE) and ''smooth late high coos'' (SL). Experiment II utilized the same vocalizations, but sorted into a high-pitched and a low-pitched class, i.e., orthogonally to the communication-relevant dimension. We found that (a) Japanese macaques learned the SE-SL discrimination faster than the pitch discrimination; (b) the reverse was true for the controls; (c) Japanese macaques showed a right-ear advantage (presumed left hemisphere advantage) for the SE-SL distinction, but not for the pitch discrimination, and (d) controls (with one exception) showed no ear advantage for either discrimination. These demonstrations of selective attention to communication-relevant parameters of conspecific vocalizations, and neural lateralization in the perception of these vocalizations, parallel similar findings in human speech perception.