Model-Based Syllabic Prediction in the Avian Auditory Pallium

PROJECT SUMMARY / ABSTRACT Sensory inputs are often incomplete or degraded by noise. To compensate, the brain uses internal models to predict what the inputs should be from context, experience, and innate biases. When phonemes are deleted and replaced with noise, listeners still report hearing continuous speech behind the occluding noise – an illusory phenomenon known as auditory induction, or restoration (AR). The restored phoneme is biased by the phonological and lexical context, and by the listener’s native language, pointing to internal representations of invariant speech patterns that are shaped through experience driving the illusory activity in place of missing auditory input. Cognitive models of speech perception suggest that restoration likely involves top-down attention and active feedback from higher-order brain areas onto lower-level auditory areas. Local neural dynamics within auditory areas, however, may be sufficient for filling in missing information based on temporal statistical regularities, encoded by means of experience-dependent Hebbian plasticity. The neural mechanism behind the encoding and activation of internal models of speech remains poorly understood, even amongst the nonhuman animals that exhibit the auditory illusion. This proposal aims to investigate how the neural dynamics during the AR illusion are dependent on the statistical regularities of auditory experience, using the zebra finch and their acoustically rich, but stereotyped songs. The proposal’s central hypothesis is that an experience- dependent internal model of conspecific song in the zebra finch auditory pallium supports the restoration of occluded song syllables. Analysis of in vivo extracellular recordings, using unsupervised manifold inference, will determine the population-level dynamics during the illusion, in the context of both naturalistic and synthetic, syntactically altered song motifs. Operant behavioral paradigms and acoustical rearing manipulation will be used to investigate how conspecific syntactic context and auditory experience affect the strength of the restoration effect. Identifying the neural mechanism behind internal model-based auditory restoration will offer insight into the broader questions of general perception, memory, and development.