An fMRI Study of Multisensory Integration of Audiovisual Speech in Autism Spectrum Disorders: Effects of Temporal Synchrony

Background: Individuals with Autism Spectrum Disorders (ASD) exhibit atypical sensory processing in multiple sensory modalities (Iarocci, 2006), including deficits in combining information across different sensory modalities (i.e., multisensory integration). One of the strongest factors influencing multisensory integration is the temporal relationship between different sensory inputs. Recently, deficits in the realm of multisensory temporal processing have been shown in ASD (Foss-Feig, 2009; Kwakye, 2011) and appear to be differentially affected with speech stimuli (Bebko, 2006). The neural substrates for the temporal integration of both low-level (Stevenson, 2010) and speech-related audiovisual stimuli (Nath, 2012) appear to be centered on the cortex surrounding the posterior superior temporal sulcus (pSTS). In addition to its central role in multisensory temporal processes, the pSTS has also been shown to differ anatomically (Levitt, 2003) and functionally (Pelphrey, 2008) in studies with ASD participants.

Objectives: Investigate the role of pSTS in multisensory temporal integration in ASD via fMRI to address the following questions:

1) How does multisensory gain seen in BOLD activation patterns differ between ASD and typically-developing (TD) participants,
2) Does the level of multisensory integration differ between social and non-social sensory inputs or between linguistic and non-linguistic sensory inputs?

Methods: Individuals with and without ASD (N=10 per group, scanning ongoing), matched for age and cognitive abilities, underwent an fMRI scan while passively viewing auditory-only, visual-only, synchronous audiovisual and asynchronous audiovisual presentations with three types of stimuli: social-linguistic (speaker reading a passage of a story), social non-linguistic (speaker producing non-speech verbal noises), and non-social, non-linguistic stimuli (hand playing notes on a keyboard). Concurrent eye-tracking ensured participants attended to the stimuli. Whole-brain GLMs were performed, and a region-of-interest analysis localized pSTS via a conjunction analysis of activations to auditory- and visual-only presentations independently for each stimulus type. BOLD responses to synchronous and asynchronous audiovisual presentations were then extracted from these ROIs and compared within and across stimulus types and groups.

Results: Data from both groups show reduced peak BOLD activation in the pSTS in response to synchronous relative to asynchronous audiovisual presentations, likely reflecting an increase in processing efficiency associated with multisensory binding (Stevenson, 2010; 2011). In the TD group, responses with social-linguistic stimuli showed the greatest reduction in peak BOLD signal (17%) relative to non-social, non-linguistic stimuli (7%). In contrast, analyses to date of our ASD cohort showed less reduction with social-linguistic stimuli (12%) than with non-social, non-linguistic stimuli (26%). Thus, our TD group showed greater processing efficiency gains with audiovisual speech, whereas the ASD group showed greater gains with audiovisual non-speech stimuli.

Conclusions: These data suggest the behavioral instantiations of impaired multisensory temporal processing seen in previous studies (Bebko 2006, Foss-Feig 2009, Kwakye 2011) may be, at least in part, a result of atypical neural processing in pSTS, particularly the impairments specific to language. Interestingly, perceptual-feedback training focused on multisensory temporal processing has been shown to improve TD individuals’ abilities to integrate audiovisual stimuli (Powers 2009) through neuro-plastic changes in the pSTS (Powers 2012) and as such may be a possible remediation tool.