Modeling Individual Differences in the Temporal Dynamics of Spoken Word Recognition in Autistic Young Adults

PROJECT SUMMARY/ABSTRACT
Autism spectrum disorder is characterized by impairments in social-emotional reciprocity and restricted
behaviors and interests. Language outcomes in autism are heterogenous; impairments in language appear early
in development, are associated with poor functional outcomes, and are resistant to treatment. Individual
differences in language abilities are associated with language efficiency; decreased efficiency in spoken word
recognition contribute to language impairments in autism, though to date the mechanisms are unspecified.
Studies of clinical populations suggest two candidate mechanisms: inefficient competition suppression and
slowed auditory processing. The proposed predoctoral training and research, which addresses a top NIDCD
research priority, will utilize EEG and machine learning to examine individual differences in the temporal and
neural dynamics of spoken word recognition in autism, and their relationship to standardized measures of
language abilities. Autistic adults ages 18-35 with a wide range of language abilities, and language-matched
neurotypical (NT) peers (n = 25 autistic, 25 NT) will complete an EEG/ERP spoken word and nonword recognition
design. Computational modeling will employ a support vector machine classification framework using ERP
response profiles to decode what word an individual heard. This “decoder” captures individual differences in
word recognition precisely and reliably. Analyses will use machine learning to achieve the following Specific
Aims and test associated hypotheses: (1) Model group-level ERP responses to words and nonwords in autistic
versus NT groups; (2) Test individual decoder accuracy as a predictor of language abilities; and (3) Test decoder
accuracy for words embedded in sentences of varying linguistic predictability. The ERP paradigm presents
minimal task demands, and is well suited for studies of individuals with language and cognitive impairments.
Theoretically driven aims and hypotheses will clarify the role of efficiency in spoken word recognition in language
deficits in autism to inform targeted early intervention. Training Plan: The individually tailored training plan
dovetails with research activities and includes methodological training in neuroimaging and computational
modeling; education in language, autism, cognitive neuroscience, and responsible conduct of research; and
professional development in scientific and community settings. Proposed activities include coursework,
interdisciplinary mentorship, methodological workshops, professional seminars, community outreach,
conference presentations, and manuscript preparation. Environment: The mentorship team has a strong NIH
funding record. Univ. of Connecticut is a Research 1 public institution, and the Dept. of Psychological Sciences
ranks in the top 10 for U.S. grant funding. UConn boasts a large community of interdisciplinary language sciences
researchers and a research-dedicated neuroimaging center that prioritizes graduate training. The application is
ideally situated to accomplish the current fellowship proposal.

Grant Number: 5F31DC022187-02
NIH Institute/Center: NIH
Principal Investigator: Rebecca Canale