Finding motifs in birdsong data in the presence of acoustic noise and temporal jitter.

Here we present a novel approach to quickly and reliably find long (200 ms - 2 s) stereotyped sequences of sounds ("motifs") in acoustic recordings of birdsong. Robust and time-efficient identification of such sequences is a crucial first step in many studies ranging from development to neuronal basis of motor behavior. Accurately identifying motifs is usually hindered by the presence of animal-intrinsic variability in execution and tempo, and by extrinsic acoustic noise (e.g., movement artifacts, ambient noise). The algorithm we describe in this report has been optimized to work in bird species that sing stereotyped syllable sequences (such as the zebra finch), and requires minimal user involvement (∼ 5 min for over 1,000 motifs). Importantly, it is transparent and robust to the choice of parameters.