Link between flow and performance updated with physiological markers

Flow is an intrinsically motivating psychological state generated by the exercise of skilled control in a demanding activity. It is known to be highly interconnected with increased performance and lowered self-criticism [1]. Cowley et al [2] tested the hypothesis that the relationship between Flow and performance may not be driven by the absolute performance, but instead by the discrepancy between anticipated and observed performance outcomes (the so-called Flow deviation, or F~d model); an effect which scales with task experience as anticipation of performance becomes more precise. The explanatory power of this model is strong but not perfect. Here, we examine how psychophysiological variance can help explain the remaining variance.
Flow state is typically assessed through questionnaires, which are susceptible to biases due to subjective self-reporting. Similarly, anticipated performance is derived from projections based on past performances. While actual performance can be measured, its perception by participants remains difficult to quantify. In light of these challenges, we employed a data-driven methodology to identify psychophysiological markers that could contribute to bridging the gap in the F~d model.
We analyzed longitudinal data from 18 participants. They performed a steering task designed to reliably induce Flow by balancing skill and demand levels. During the sessions, we recorded a set of physiological signals from participants, including gaze positions, pupil sizes, blood volume pressure (BVP), and electrodermal activity (EDA). After each trial, participants were asked to fill out a Flow assessment questionnaire.
From the gathered signals, a set of descriptive features were extracted, based on the first ten seconds of each trial to control for behavioural variance. Linear mixed models of Flow predicted by performance deviation were tested with added fixed effect of psychophysiology, adding one such features to each model to find which explained the most Flow variance.
By incorporating physiological features into the F~d model, we observed an increase in the explained variability for certain features. The most significant of which is gaze transitional entropy, a measure of complexity of gaze behavior. Features related to EDA and BVP did not significantly explain the variance in Flow.
Neither of the markers associated with arousal states demonstrated a significant contribution to F~d model. In addition, gaze transitional entropy may manifest enhanced peripheral vision, usually associated with relaxed states. While Flow is commonly linked to a high arousal state, our results suggest that it is not the only prerequisite. Our findings underscore the need for further exploration into the combined effects and interplay of neural and cognitive components of Flow.