MENTAL WORKLOAD AND PERFORMANCE DYNAMICS UNDERLYING HIERARCHICAL PLANNING PROCESSES DURING COMPLEX ACTION SEQUENCE EXECUTION UNDER VARIOUS DEMANDS

Abstract

Mental workload arises from the link between task demands and the engagement of processing resources. While mental workload has been studied extensively, only a limited body of work has examined this notion as it relates to the generation of discrete, goal-oriented action sequences which are critical to most daily activities (e.g., making a cup of coffee). Such action sequences engage high-level (cognitive) processes to plan the order of actions in combination with low-level (sensorimotor) processes for movement execution. The study of this hierarchical mechanism is important to inform how action sequences governed by a goal and rules are generated such that some are more parsimonious (e.g., fewer actions) than others. In such a context, the examination of mental workload could further characterize the engagement of high- and low-level processing resources along with their interaction to face cognitive and sensorimotor demands. Although prior efforts have examined discrete action sequence performance, they did not study mental workload dynamics via a combined assessment of complementary neurophysiological correlates (i.e., regional activity, functional connectivity) while simultaneously manipulating both cognitive and motor demands to characterize the relationship between high- and low-level processing resources. Further, most prior action sequence studies overlooked the functional interaction between regions thought to be critical for coordinating these high- and low-level processes such as the dorsolateral prefrontal (dlPFC), premotor (PMC), and sensorimotor (SMC) cortices. Also, most of these studies were clinical without considering healthy individuals or computational tools that can assess in detail the structure of the performed action sequences. Therefore, through a combined evaluation of performance and brain dynamics, this work aimed to examine the engagement of high- and low-level processing resources underlying mental workload when individuals execute action sequences under varying cognitive and motor demands. In the present study, participants solved an action sequence task (Tower of Hanoi) under normal (i.e., physical) and altered (i.e., virtual) sensorimotor conditions and low and high cognitive demand. Performance was examined using task completion time and sequence structure (Levenshtein distance). Regional EEG power activity along with functional connectivity between the dlPFC, PMC, and SMC were used to evaluate cortical effort and cortico-cortical communication in the theta, low-alpha, and high-alpha bands. Sensorimotor mu-rhythm, frontal theta/parietal alpha (FTPA), and frontal theta/frontal alpha (FTFA) power ratios were also computed. Results revealed larger task completion time, elevated EEG power in all three frequency bands in the anterior-frontal and occipital regions, enhanced temporal theta and low-alpha power, and attenuated mu-rhythm during the normal relative to the altered sensorimotor conditions. Cortico-cortical communication between the dlPFC and both the PMC and SMC increased with cognitive demand under the normal but not altered conditions. Further, larger task completion time and Levenshtein distance were associated with enhanced theta power in the anterior-frontal, frontal and occipital regions, increased FTFA ratio, and larger functional theta connectivity as cognitive demand increased. Findings indicate degraded performance along with increased mental workload as revealed by greater recruitment of cognitive-motor resources in response to increased cognitive demand and under normal relative to the altered conditions suggesting that the former imposed larger sensorimotor demands compared to the latter. Connectivity results suggest resources may, to some extent, be shared when facing combined elevation of cognitive and motor demands and highlight the PMC’s role in the low- and high-level interface for action sequence execution. These results may inform applications related to physical and/or virtual execution of action sequence tasks such as, but not limited to, rehabilitation, human factors, and human-robot interactions.