An Operantly Conditioned Looking Task for Assessing Infant Auditory Processing Ability
Jason Nawyn, Cynthia Roesler, Teresa Realpe-Bonilla, Naseem Choudhury, April A. Benasich · 2007 · Proceedings of the 9th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '07) · doi:10.1145/1296843.1296869
Summary
This paper from the Center for Molecular & Behavioral Neuroscience at Rutgers University describes the design and evaluation of a gaze-driven interface for assessing rapid auditory processing (RAP) abilities in infants aged 4 to 6 months. The research is motivated by evidence that deficits in RAP — the ability to process rapidly changing auditory signals occurring within tens of milliseconds — are characteristic of individuals with language-based learning impairments (LLI), affecting roughly 20% of preschool and school-aged children, with over 50% going on to develop dyslexia. Critically, longitudinal studies show that infant performance on RAP tasks at 16, 24, and 36 months predicts later language outcomes, and early intervention during the first year of life may be maximally effective. The system uses a cross-modal operant conditioning procedure: infants hear a continuous stream of familiar auditory stimuli (pairs of 800 Hz tones), and when a novel stimulus (800 Hz + 1200 Hz tone pair) is introduced, the infant must learn to anticipate a video reward by directing their gaze to a specific display area before the reward appears. An automated eye tracker (Applied Science Laboratories Model 504) monitors the infant's gaze, while the system manages stimulus presentation, reward delivery, and response scoring through a three-phase protocol: training, criterion, and testing.
Key findings
Of 21 infants enrolled, 18 completed the evaluation (mean age 6 months 2 days). In the training phase, 89% of participants passed (learned the stimulus-reward association). In the criterion phase, group performance was significantly better than chance (t(16) = -8.796, p < 0.001), with 11 of 16 (69%) who passed training performing significantly above chance individually. When the traditional 4/5 passing rule was applied, 100% of trained infants demonstrated passing performance. The complete experimental session averaged only 11 minutes from seating to completion. However, several important design challenges emerged: eye tracking in infants reliably captures gaze only about 50% of the time under typical conditions (versus 90%+ in optimal conditions), requiring a human coder as backup. Misleading environmental cues proved problematic — a chime sound paired with the video reward was more salient than the auditory stimulus and had to be removed, and spatial localisation of the audio stimulus interfered with learning the cross-modal association. The fixation point design required careful balance: engaging enough to maintain attention but not so stimulating as to compete with the auditory task. Simple animated geometric figures worked better than faces or complex designs.
Relevance
This paper represents an innovative intersection of HCI, clinical neuroscience, and early intervention for developmental disabilities. The core accessibility contribution is the development of a technology-mediated clinical tool that could enable identification of children at risk for language learning impairments far earlier than current methods allow — potentially before a child's first birthday, when intervention may be most effective. For accessibility practitioners, the paper offers valuable lessons about designing interfaces for non-verbal, pre-linguistic users who cannot be instructed or provide explicit feedback. The design challenges encountered — competing sensory cues, attention management, unreliable eye tracking with young infants, and the need for sessions short enough for infant tolerance — are relevant to anyone designing technology for users with significant cognitive or communication disabilities. The research also highlights how assistive assessment tools can shift the paradigm from reactive intervention (after a disability is diagnosed) to proactive prevention (identifying risk factors early enough to alter outcomes).
Tags: early intervention · eye tracking · infant assessment · auditory processing · language impairment · operant conditioning · clinical tools · developmental disabilities · dyslexia