Abstract
Tactile cues serve as a vital alternative, or complement, to visual and auditory stimuli in human-machine interfaces. The ability to accurately detect changes in tactile stimuli is therefore fundamental to the success of these interfaces. However, the phenomenon of tactile change blindness—where tactile changes go unnoticed—poses a significant risk to safety in complex domains, such as aviation and driving. Despite its importance, the mechanisms underlying tactile change blindness remain poorly understood, complicating efforts to optimize vibrotactile cueing. While change detection has been characterized at the neural level in vision and audition, little is known about the neural events underlying tactile change detection.
This dissertation addresses these challenges by integrating neurophysiological techniques with behavioral measures to characterize both tactile change detection and change blindness across vibrotactile cues defined by spatial and featural properties—specifically, changes in location, intensity, pulse number, and pulse width. Across three experiments, seventy-two participants completed a tactile change detection paradigm, where they were tasked with determining whether a tactile change had occurred. Meanwhile, their brain activity between post-change stimulus onset and their response was captured using electroencephalography (EEG) and analyzed using the event-related potential (ERP) technique.
Intensity and location cues produced the highest change detection accuracy, yet a marked advantage in reaction time was observed for all featural cues compared to changes in location. At the neural level, several ERP components, including the N80, P1, N140, and P3, were identified as the electrophysiological correlates of tactile change detection and change blindness. For the first time, these components were mapped onto five theoretically derived stages of the change detection process. The P3 was most critical to tactile change detection, reflecting comparison of pre- and post-change stimulus representations in the working memory and the identification of discrepancies between them. Under distracting conditions, neural indicators of increasing tactile perceptual load were identified, including the P45, which appeared selectively under distraction.
The electrophysiological data captured were then leveraged to develop time-resolved decoding models, which identified time windows critical to predicting (a) the type of tactile stimulus presented to human subjects and (b) the subject’s response. Time windows aligned with P3 onset were most informative for predicting both stimulus type and participant response. Ultimately, complex, multichannel brainwave data was transformed into an objective record of tactile sensory experience, equipped with the ability to predict when someone will achieve tactile change detection.
The collective findings of this dissertation advance our understanding of tactile change detection, with broad implications for the design of tactile, multimodal, and adaptive displays, as well as brain-computer interface technologies. Prevailing assumptions about the role of the primary somatosensory cortex in higher-order cognitive processes, like perception and decision-making, are challenged. By integrating behavioral, neurophysiological, and modeling outcomes, this work lays the groundwork for the thoughtful integration of vibrotactile cues into human-machine interfaces, and for anticipating when a tactile change might go unnoticed.