Cognitive Load Dynamics in Training: Analysis and Insights
SOMAREALITY
Introduction
Training is critical for improving skillsets across domains like sports, aviation, and education. Trainees must balance challenging tasks with processing information and feedback, which induce cognitive loads. Optimal training demands strike a balance: tasks must be challenging enough to drive learning without overwhelming information processing capabilities.
Eye tracking offers real-time insight into learning processes by monitoring task-evoked pupillary responses. This study utilizes SOMAREALITY’s eye-tracking-based Cognitive Load (CL) biomarker to track cognitive processing during a ball-interception task. The core hypothesis is that optimal learning features strong task-evoked CL spikes followed by a recovery to lower baseline levels, whereas overload is marked by high CL peaks that lack sufficient recovery.
Method
The study analyzed 29 medical university participants (18 female, aged 21–43) completing a computer-based ball-interception task that taps into visuomotor skills. Participants predicted the motion path of a parabola-launched ball obscured mid-flight, clicking as close as possible to its landing point before screen-edge exit. Difficulty was intentionally set high initially without practice opportunities. Participants received post-response visual positioning feedback and Euclidean distance error scores.
The experiment consisted of 4 blocks of 40 trials each, featuring varying ball velocities and trajectories. A 20-minute break involving a neutral nature documentary occurred after block 2, and numeric feedback was removed following block 3 to manipulate learning rates. Eye-tracking glasses (Tobii X or Pupil Neon) recorded pupillary data, and logarithmic learning curves were fitted per block to quantify error dynamics and learning rates. Peak CL values were correlated with errors and learning rates.
Results
Across participants, average CL peaked strongly at 2.10 seconds post-stimulus, significantly higher than baseline motion onset levels. Peak timing varied individually from 1.24 to 2.59 seconds.
Logarithmic learning curve fits were moderate. Learning rates varied significantly by block and individual participant. Crucially, block-by-block mean CL peaks showed a positive correlation with learning rates, whereas trial-by-trial CL peaks showed only a negligible correlation with Euclidean error.
Comparing the top and bottom 10% of performances revealed that CL peaked significantly higher during trials associated with high learning rates and high Euclidean errors. However, unbaselined (raw) analyses showed that average absolute tonic CL was significantly elevated for the lowest learning rates and highest errors.
Discussion
The 2-second post-stimulus peak indicates that cognitive processing extends well beyond the half-second ball-motion window into the post-trial phase, confirming that eye-based CL captures deep cognitive engagement rather than mere visual processing. The robust correlation between CL peaks and learning rates underscores the active role of ongoing cognitive processing in skill acquisition.
These findings refine the “optimal load” training framework. While optimal learning and overload can generate identical peak load magnitudes, optimal conditions allow cognitive load to recover downward, whereas overload conditions leave CL chronically elevated. Real-time CL monitoring can help coaches precisely manage rest intervals between challenging tasks to prevent resource depletion and maximize retention.
References
Bosco, G., Delle Monache, S., & Lacquaniti, F. (2012). Catching What We Can’t See: Manual Interception of Occluded Fly-Ball Trajectories. PLoS ONE, 7(11), e49381. https://doi.org/10.1371/journal.pone.0049381
Chen, O., & Kalyuga, S. (2019). Cognitive load theory, spacing effect, and working memory resources depletion: Implications for instructional design. Loughborough University. https://doi.org/10.4018/978-1-5225-9833-6.ch001
Cheng, N., Gao, C., Yang, M., Wu, Y., Wang, G., & Xiao, C. (2014). Analysis of the learning curve for beating heart, totally endoscopic, coronary artery bypass grafting. The Journal of Thoracic and Cardiovascular Surgery, 148(5), 1832-1836. https://doi.org/10.1016/j.jtcvs.2014.02.002
Guadagnoli, M. A., & Lee, T. D. (2004). Challenge point: A framework for conceptualizing the effects of various practice conditions in motor learning. Journal of Motor Behavior, 36(2), 212-224. https://doi.org/10.3200/JMBR.36.2.212-224
Harris, D. J., Arthur, T., Vine, S. J., Liu, J., Abd Rahman, H. R., Han, F., & Wilson, M. R. (2022). Task-evoked pupillary responses track precision-weighted prediction errors and learning rate during interceptive visuomotor actions. Scientific Reports, 12(1), 22098. https://doi.org/10.1038/s41598-022-26544-w
Radvansky, G. A., Parra, D., & Doolen, A. C. (2024). Memory from nonsense syllables to novels: A survey of retention. Psychonomic Bulletin & Review, 31(6), 2437-2464. https://doi.org/10.3758/s13423-024-02514-3
Ramsay, C. R., Wallace, S. A., Garthwaite, P. H., Monk, A. F., Russell, I. T., & Grant, A. M. (2002). Assessing the learning curve effect in health technologies. Lessons from the nonclinical literature. International Journal of Technology Assessment in Health Care, 18(1), 1-10.
