Flow Experiences During Visuomotor Skill Acquisition Reflect Deviation From a Power-Law Learning Curve, but Not Overall Level of Skill

Abstract

Flow is a state of "optimal experience" that arises when skill and task demands match. Flow has been well studied in psychology using a range of self-report and experimental methods; with most research typically focusing on how Flow is elicited by a particular task. Here, we focus on how the experience of Flow changes during task skill development. We present a longitudinal experimental study of learning, wherein participants (N = 9) play a novel steering-game task designed to elicit Flow by matching skill and demand, and providing clear goals and feedback. Experimental design involves extensive in-depth measurement of behavior, physiology, and Flow self-reports over 2 weeks of 40 game trials in eight sessions. Here we report behavioral results, which are both strikingly similar and strong within each participant. We find that the game induces a near-constant state of elevated Flow. We further find that the variation in Flow across all trials is less affected by overall performance improvement than by deviation of performance from the expected value predicted by a power law model of learning.

Keywords: Flow; high performance cognition; power law of practice; skill acquisition; steering; visuomotor performance.

Figure 1

The high-speed steering task. The participant steers the blue cube to avoid conical/spherical obstacles on the track, which is bounded to each side by dark blue parallel lines. The game was designed to continually adapt the difficulty level (speed) to the participant's skill (obstacle collisions). Such balance is considered one of the key antecedents of Flow.

Figure 2

The game was played in eight sessions on eight different days. In sessions 1 and 5–8, physiological signals were recorded during task performance; in sessions 2–4(*) no physiology was recorded. Each session consisted of five trials (2–4 min) followed by a self-report questionnaire (FSS, Flow Short Scale) about the latest trial.

Figure 3

(A) Participant-wise data showing logarithm-transformed performance and Flow self-reports in the speeded steering task. Ordinate shows log-duration of trials, abscissa shows log-cumulative trial count. Dashed blue lines fitted to the data are power-law learning curves, which transform to linear in log-log space. 95% confidence intervals around the slope are grayed. Flow scores (z-scored) are indicated by color. (B) Participant-wise deviation scores (observed trial duration minus predicted trial duration) plotted against Flow scores for each participant, and fitted by linear models. 95% confidence intervals around the slope are grayed.

Figure 4

Violin plot representing participants' self-reported Flow in sessions 1–8 (per-session Flow = mean of five trials. The self-report items are given in Supplementary Information, the scale was 1–7).

Figure 5

Duration and flow over sessions grouped: 1 (Introduction+physiological measurements), 2–4 (training), 5–8 (physiological measurements), N = 72.