Using video games to study the acquisition and performance of psychomotor skills

Understanding how humans learn complex skills is a fundamental aim of cognitive science. Digital games offer promising opportunities to study cognitive factors associated with skill acquisition and performance, as they motivate longitudinal engagement and produce rich, multivariate data sets. By applying mutlivariate analysis techniques to data arising from gameplay, this thesis extended the literature on cognition as it pertains to psychomotor skill. We describe three studies
that were conducted in this regard. In the first study, we analyzed the relationship between the temporal distribution of play instances and performance in a commercial digital game (League of Legends). Using clustering techniques and big data, we demonstrated that players who cram gameplay into short time frames ultimately perform worse than those who space the same number of games over longer periods. In the second study, we examined an experimental data set of participants who played Meta-T, a laboratory version of Tetris. Using Principal Components Analysis and regression techniques, we identified cognitive-behavioural markers of performance, such as action-latency and motor coordination. We also applied Hidden Markov models (HMM) to time series of these markers, showing that moment-to-moment dynamics in performance can be segmented into behavioural states related to latent psychological states. In the third study, we investigated the neural correlates of behavioural states during performance. Using simultaneous MEG and behavioural recordings of participants playing Tetris, we segmented time series datasets of neural activity based on time stamps of behavioural epochs derived from HMMs. We compared behavioural epochs based on neural markers, showing that cognitive states derived from multivariate behavioural data correlate with neural activity in the alpha band power. Taken together, this thesis advances our understanding of using digital game data to study cognition and learning. It demonstrates the feasibility of recording high-density neuroimaging data during complex behavioural tasks and obtaining reliable measures of internal neuronal states during complex behaviour.