A Tradeoff Between Performance and Accuracy in Bird Song Learning


David C. Lahti, Department of Biology, Queens College, City University of New York, 65-30 Kissena Blvd., Flushing, NY 10541, USA.
E-mail: david.lahti@qc.cuny.edu


Some of nature’s most complex behaviors, such as human speech and oscine bird song, are acquired through imitative learning. Accurate imitative learning tends to preserve patterns of behavior across generations, thus limiting the scope of cultural evolution. Less well studied are the routes by which cultural novelties arise during development, beyond simple copy error. In this study we assess, in a species of songbird, the relationship in song learning between two potentially conflicting learning goals: accuracy in copying and maximization of vocal performance. In our study species, the swamp sparrow (Melospiza georgiana), vocal performance can be defined for a given song type and frequency range by the rate of note repetition (‘trill rate’), with faster trills being more difficult to sing. We trained young swamp sparrows with song models with experimentally modified trill rates and characterized both the accuracy and performance levels of copies. Our main finding is that birds elevated the trill rates of low-performance models, but at the expense of imitative accuracy. By contrast, birds reproduced normal and high-performance models with typically high accuracy in structure and timing. Developmental mechanisms that enable songbirds to balance imitative accuracy and vocal performance are likely favored by sexual selection and may help explain some current patterns of variation in birdsong. Such mechanisms may also explain how behaviors that are learned by imitation can nevertheless respond to selection for high-performance levels in their expression.