The impact of noise on temporal integration of speech in the human brain

PROJECT SUMMARY
Understanding speech in real-world conditions is a complex process that requires the brain to integrate
information about the incoming speech stream concurrently on multiple timescales, ranging from milliseconds to
seconds. While previous work has characterized integration timescales across the auditory cortex, it remains
unclear the extent to which these temporal integration windows are fixed or whether they vary depending on
stimulus processing demands, such as the presence of background noise. Prior studies that have examined this
question have been limited to measuring integration windows using linear modeling (e.g., spectrotemporal
receptive fields), and much of the relevant research has either been conducted in animals or used coarse
neuroimaging measures. As a consequence, much remains unknown about the human auditory cortex integrates
information in speech during challenging listening conditions, which is thought to depend upon highly nonlinear
computations. In this project, we examine the degree to which auditory cortical integration windows vary
depending on the presence or absence of background noise using a novel method (the “temporal context
invariance” or TCI paradigm) applied to both scalp EEG (Aim 1) and intracranial EEG recordings (Aim 2). The
TCI paradigm makes it possible to measure integration windows from any sensory response, even if that
response is a highly nonlinear function of its input. Scalp EEG recordings will allow me to test if there is any
overall change in the integration window of auditory cortical responses in the presence of noise, while the
unparalleled spatiotemporal resolution of intracranial recordings will enable me to examine the neuroanatomical
basis of integration window flexibility. The proposed research will answer longstanding questions about the
nature of temporal integration in the auditory cortex, and further our understanding of how the brain reckons with
the extreme variability inherent in real-world communication settings in order to arrive at stable representations
of speech despite interference from background sounds. This research is a critical first step in understanding the
speech perception deficits in noise that are present in auditory neurodevelopmental and attentional disorders,
many of which are hypothesized to also involve impairments in temporal processing. In the process of conducting
this research, I will develop expertise in several valuable domains: (1) scalp EEG experiments, (2) intracranial
EEG experiments, (3) the analysis of high-dimensional time-series data, (4) hypothesis-driven encoding models
of speech. These skills complement my prior expertise in fMRI, music, and data-driven component modeling,
thus equipping me with a unique and valuable set of experimental and computational skills that will facilitate my
transition to an independent research career.

Grant Number: 5F32DC022145-02
NIH Institute/Center: NIH
Principal Investigator: Dana Boebinger