Manipulating image luminance to improve eye gaze and verbal behavior in autistic children

Decades of research have identified atypical global processing as a core feature of autism, often attributed to local interference, weak central coherence, enhanced perceptual functioning, or differential use of spatial frequency. Existing theories offer varied mechanisms ranging from stimulus saliency to executive functioning, but interventions directly impacting the lives of autistic individuals are lacking. This study addresses this gap by focusing on the link between visual attention and semantic representation. The research questions are: 1) Does manipulating image luminance and spatial frequency increase fixations in areas of interest (AOIs) for autistic children? 2) Does manipulating these image characteristics increase the likelihood of global verbal responses? The study aims to leverage the relationship between sensory processing and social communication by developing a digital filter to offload low-level sensory processing of global features, thereby improving visual attention and semantic communication.

The literature review extensively examines existing theories on atypical global processing in autism, including local interference, weak central coherence, and the spatial frequency hypothesis. It highlights the mixed findings across studies and the need for interventions that directly improve global processing and social communication. The review also discusses the challenges of measuring eye gaze behavior in children and the use of eye tracking as a measure of cognitive processes. Previous research on priming and cognitive control in autism is discussed, emphasizing the importance of both automatic and purposeful executive functions. The limitations of current methods in inducing and measuring processing of stimuli are mentioned. Finally, previous work on visual processing differences in autism, focusing on temporal binding, processing biological motion, and local interference in static stimuli, is reviewed, highlighting the need for interventions at the level of static stimuli.

This study employed a 2 x 2 within-group factorial design (baseline x filter, session x 2), using a think-see-say paradigm with 11 autistic children (ages 9-18). Participants viewed 50 images in both baseline and filtered conditions across two sessions. Eye gaze and verbal responses to the prompt “What is this picture about?” were recorded. The global filter reduced image characteristics found to be distracting for autistic individuals by desaturating and blurring non-relevant details based on neurotypical gaze patterns from an open-source dataset (OSIE). Verbal responses were scored using a rubric with three levels: 0 (incorrect/unrelated), 1 (irrelevant/local details), and 2 (plausible global description). Eye gaze data was analyzed to determine fixation rates in AOIs. Image characteristics analyzed include luminance, spatial frequency, hot spot size, and semantic content (living vs. nonliving). A general linear regression analysis was conducted to investigate the relationships between global processing measures (verbal responses and eye fixations), participant characteristics (age, social pragmatic language impairment), and image characteristics.

The analysis revealed several significant findings. Age significantly impacted both fixations in hot spots and global verbal responses; older participants showed higher likelihoods of both. The baseline condition increased the likelihood of fixations in hot spots, while the filtered condition increased the likelihood of global verbal responses in the second session. Images in the second session, regardless of condition, resulted in more fixations in hot spots. The semantic level of the image (living vs. nonliving) significantly predicted global verbal responses, with living images eliciting more global responses. Lower luminance in non-salient areas (cold spots) increased the likelihood of fixations in hot spots, whereas higher luminance in cold spots increased the likelihood of global verbal responses. The interaction between luminance and spatial frequency was significant for fixations in both hot and cold spots, with higher luminance impacting the likelihood of hot spot fixations. The study achieved good inter-rater reliability for both verbal responses (87%) and eye tracking data (82%).

The findings suggest that manipulating image characteristics, particularly luminance, can positively influence both eye gaze and verbal behavior in autistic children. The contrasting effects of luminance in cold spots on fixations and verbal responses highlight the complexity of visual attention and suggest a need for further investigation into the relationship between these two measures. The study's limitations include the potential ambiguity of the verbal prompt, the significant session order effect, and the limited sample size. However, the results support the feasibility of augmenting global processing through visual stimulus manipulation and provide valuable insights for designing assistive technologies. The importance of considering individual differences, such as age and language ability, is underscored. The findings also suggest the need for future studies using longer sessions to examine the stability of eye gaze behavior and exploring different approaches to luminance manipulation.

This study provides initial evidence that manipulating image characteristics, especially luminance, can be a viable strategy for improving global processing in autistic children. The development of a digital filter based on these findings could lead to innovative assistive technologies to enhance social communication. Future research should focus on optimizing the filter design, expanding the sample size, and conducting longitudinal studies to evaluate the long-term effects of such interventions. This research opens new avenues for understanding and supporting visual attention and social communication in autism and potentially other neurodevelopmental disorders.

The study's limitations include the potential ambiguity of the verbal prompt, which could have been challenging for younger and less verbally fluent participants. The significant session order effect suggests that repeated exposure to stimuli or a learning effect influenced the results. The relatively small sample size limits the generalizability of the findings. The study also did not directly measure local interference, instead relying on previous research indicating its prevalence in the population sampled. Further research is needed to address these limitations and explore the wider applicability of these findings.