Determining how information flows within neural pathways is a fundamental requirement for understanding how the brain perceives, learns, and produces behavior, and how neural disorders disrupt normal cognitive processing. The goal of this proposal is to address this requirement by developing a publicly available set of inference algorithms for deciphering information flow networks in the brain, using the songbird vocal communication system as our testing and experimental model. Songbirds are one of the only accessible non-human animals where learned vocal communication, the substrate for human language, can be studied. Non-human primates, rodents, and other commonly studied animals do not have this ability. Using multielectrode array electrophysiology data from songbirds, the PIs and their colleagues have recently developed new computational methods for automatically inferring models of neural flow networks. Here, electrophysiology data will be collected from the auditory and vocal pathways of the songbird brain and provided to our new computational inference methods, in order to automatically infer the architecture of the neural flow networks that arise and evolve during auditory and vocal learning. With this tool, we can study the network mechanisms of neural deterioration of learned vocalization that occurs due to deafening, as in humans; we can study basal ganglia disorders that affect learned communication-including stuttering- thus gaining better insight into treating auditory, vocal, and mental disorders. The results will allow the PIs to formulate and experimentally test models that describe how vocal learners learn to recognize and vocalize the sounds they hear in their environment. The proposed methods will enable investigators to monitor-in near real time-how different regions of the brain exchange information during various processing and learning tasks. The proposed algorithms are applicable to a wide range of biological (and non-biological) problems. Any neural system can be studied and the methods can be applied to a range of different types of data, e.g. single or multi-unit recordings, EEG, or fMRI. To facilitate this, an entire aim of the proposal is dedicated to producing user-friendly software of high quality and modularity for general use throughout the neuroscience and general science communities. This will enable scientists to better understand how neural anatomical networks are actively utilized and how that utilization evolves over time.
| Smulders, Tom V; Jarvis, Erich D (2013) Different mechanisms are responsible for dishabituation of electrophysiological auditory responses to a change in acoustic identity than to a change in stimulus location. Neurobiol Learn Mem 106:163-76 |
| Smith, V Anne; Yu, Jing; Smulders, Tom V et al. (2006) Computational inference of neural information flow networks. PLoS Comput Biol 2:e161 |
