UC San Diego

The adaptive value of learning is constrained by generalization. Whether an organism successfully generalizes knowledge carries lofty implications for such things as navigating new paths, employing foraging innovations and identifying conspecifics. Generalizing knowledge of patterned relationships, however, is considered a hallmark of human faculties, such as music, analogical reasoning and language. In the auditory domain, generalizing knowledge of sound-based patterns is thought to underlie important computational processes necessary for language acquisition. In this dissertation, we take a comparative approach to consider the scope and constraints of auditory pattern generalization in a non-human species, the European starling songbird. We begin by establishing the ecological validity of studying pattern generalization in starlings by showing that the sequential patterning of components (motifs) in male starling songs carries information about likely relative fitness. Crucially, female starlings are sensitive to the sequential patterning of motifs in male song, evinced by strong preferences for some pattern types over others. From here, we provide definitive evidence that starlings can learn and generalize complex patterns built from natural auditory communication signals in controlled operant settings. We go on to identify key constraints underlying their ability to generalize patterns. These findings offer important insights into the relationship between the structure of a pattern and its constituents in real-world signals and how this relationship impacts the ability to generalize. We propose that the pattern generalization abilities of starlings resemble critical processes observed during language acquisition in human infants, such as auditory rule learning. We conclude that starlings constitute an important model organism for language- relevant learning recent advances in the behavioral and neural basis of temporal and syntactic processing of natural auditory communication signals in the starling