Complex vocal behavior and cortical-medullar projection K. Okanoya(1,2), S. Hihara(3), N. Tokimoto(1), Y. Tobari(1), & A. Iriki(3). 1. Chiba Univ., 2. Japan Sci. & Tech. Agency, 3. Tokyo Med. & Dent. Univ. okanoya@cogsci.L.chiba-u.ac.jp Vocal learning independently evolved several times in vertebrates. Species of whales, bats, birds, and humans exhibit vocal learning. Are there any specific anatomical substrates that correlated with the faculty of vocal learning? One of the candidates for this question is the direct cortical-medullar pathway for articulation and breathing. In humans, a part of motor cortex directly projects to the medullary nuclei, the nucleus ambiguus and the nucleus retro-ambiguus. This projection was absent in the squirrel monkey and Jurgens (2002) assumes that this projection exists only in humans. Similarly, there is a direct cortical-medullar pathway for articulation and breathing in the zebra finch, a species of songbirds, but a similar projection in pigeons that do not learn to sing (Wild, 1993). Thus, this projection exists in the species that show vocal learning while it is absent in the species without vocal learning. While this projection exist only in a limited number of species, it may be possible, that this is simply very faint in most of species. In that case, by training animals to perform spontaneous vocalization while they are young, we may be able to reinforce this and induce vocal learning in a species that was said to be non vocal learners. A report on the spontaneous vocal differentiation in Japanese macaques (Hihara et al., 2003) and that on the spontaneous construction of nested hierarchical structure in a species of rodent, Degus, are suggestive of such possibility although no anatomical data were available at present. When trained to use a rake to retrieve a distant food, monkeys began to vocalize "coo" calls spontaneously. They did so especially when the preparation of the rake tool by the experimenter delayed (Hihara et al. 2003). We systematically manipulated behavioral contexts by giving the tool or food, regardless of the types of the calls. We therefore never tried to differentiate the calls, but the monkeys spontaneously differentiated calls. The monkeys eventually used acoustically distinct types of calls when they ask for the tool or food. We argued that the different reward conditions (food or tool) set up different emotional contexts for the monkeys. Different emotional contexts, in turn, affected the production of coo calls differently for the tool or food situations. Since the tool training activated the neocortex very highly, the calls were associated with different behavioral contexts. Thus, the calls became categorized and emotionally differentiated calls gradually became categorical vocalizations. Through this process, the emotional coo calls changed into categorical labels denoting the behavioral situation. In another study, when degus (a species of rodent) were trained to vocalize in order to obtain food reinforcement, we observed they spontaneously constructed hierarchical self-embedded structure (Tokimoto & Okanoya, 2003). With a large dust-bath dish, a medium sized food cup, and a small toy ball, they spontaneously constructed a triplet "Chinese box" like structure of "the ball into the cup into the dish." This particular behavior occurred only during the period of operant training. It has been assumed that the ability to combine multiple objects hierarchically to construct self-embedded structures is restricted to primates including humans. We interpreted this as an opposite case for the monkey example. Vocal operant training probably put a heavy load to the brain of the degus and required co-activation of medullary, limbic, and cortical vocal related areas. This prepared hierarchically organized behavior in general and degus could utilize this for their "play" like, or "tool-use" like behavior. Taken together, we suspect heavy cognitive load associated with vocal behavior may prepare for the direct anatomical connection between cortical and medullary vocal centers necessary for vocal learning. To further investigate this process, we need to show how these cognitive loads could actually affect anatomical structures.