A method of inferring partially observable Markov models from syllable sequences reveals the effects of deafening on Bengalese finch song syntax

Abstract

Generative models have broad applications, ranging from language processing to analyzing bird-song. In this study, we demonstrate how a statistical test, designed to prevent overgeneralization in sequence generation, can be used to deduce minimal models for the syllable sequences in Bengalese finch songs. We focus on the partially observable Markov model (POMM), which consists of states and the probabilistic transitions between them. Each state is associated with a specific syllable, with the possibility of multiple states being associated to a single syllable. This feature sets the POMM apart from a standard Markov model, where each syllable is associated to just one state. This multiplicity suggests that syllable transitions are influenced by the specific contexts in which the transitions appear. We apply this method to analyze the songs of six adult male Bengalese finches, both before and after they are deafened. Our findings indicate that auditory feedback is crucial in shaping the context-dependent syllable transitions characteristic of Bengalese finch songs. SignificanceGenerative models are adept at representing sequences where the order of elements, such as words or birdsong syllables, depends on the context. In this study, we demonstrate that a probabilistic model, inspired by neural encoding of song production in songbirds, effectively captures context-dependent transitions of syllables in Bengalese finch songs. Our findings indicate that the absence of auditory input, as seen in deafened finches, diminishes these context dependencies. This implies that auditory feedback is vital for establishing context-based sequencing in their songs. Our method can be applied to various behavioral sequences, offering insights into the neural underpinnings that govern statistical patterns in these sequences.