PROJECT SUMMARY The ability of young learners to make successful predictions about the effects of their immature behavior on parents may promote communicative development. One mechanism by which infants? predictive skills could develop is curiosity-driven learning (CDL), which proposes that learning is directed through the intrinsic reward of reducing uncertainty via exploratory behaviors (Oudeyer & Smith, 2016). Via exposure to the probabilistic structure of the world (e.g. in visual sequences), infants may generate predictions about the probabilities of later events. Only some of these predictions will be correct, and the discrepancy can attract infant attention, with interest highest at moderate discrepancy levels (Kinney & Kagan 1976). Internally generated predictions about events may thus direct infant attention to stimuli, and shape subsequent learning. However, while CDL has been studied in non-social domains, it has not been applied to learning in social interactions. Parents may scaffold the communicative development of infants by providing interactions that adapt to infant learning progress. The aims of this project are to investigate whether CDL is a mechanism of socially guided vocal and communicative development, both in a model species, the zebra finch (Taeniopygia guttata), and in human infants. There are many parallels between zebra finches and human infants in neural and social structure, and in the role of contingency in socially guided vocal learning. Zebra finches also have a short developmental time period and an easily obtained learning outcome measure (adult song), and thus serve as an ideal model species for investigations into the mechanisms of communicative development under conditions of maximal experimental control. Findings from songbirds can also inform the human infant learning literature by indicating manipulations of response content and variation that could influence the structure of mature vocalizations. The proposed research consists of two phases carried out in both songbirds and human infants. The first phase will titrate out the influence of predictability on learning by manipulating the pattern of contingent responses to infant vocalizations, thus testing the conditions when CDL facilitates attention and learning. In zebra finches, this will be done via video playback of adult response to juveniles, and in humans, experimenter-cued parental responses to their infants. Infants of both species are expected to learn better with intermediate amounts of predictability in parental response. The second phase will observe in naturalistic contexts whether parents exhibit sufficiently predictable patterns of response to infant vocalizations that CDL could serve as a mechanism of vocal learning, and if parent characteristics predict their response consistency. Parents with more consistent (i.e. predictable) response patterns should promote infant communicative development. The proposed research, by investigating the predictive capacities of young learners (as driven by CDL) and the typical amount of predictable adult behavior in social interactions, could inform intervention programs for children with autism, and suggest how disorders in prediction may emerge over development.