Examining the Roles of Reasoning and Working Memory in Predicting Casual Game Performance across Extended Gameplay

Abstract

The variety and availability of casual video games presents an exciting opportunity for applications such as cognitive training. Casual games have been associated with fluid abilities such as working memory (WM) and reasoning, but the importance of these cognitive constructs in predicting performance may change across extended gameplay and vary with game structure. The current investigation examined the relationship between cognitive abilities and casual game performance over time by analyzing first and final session performance over 4-5 weeks of game play. We focused on two groups of subjects who played different types of casual games previously shown to relate to WM and reasoning when played for a single session: (1) puzzle-based games played adaptively across sessions and (2) speeded switching games played non-adaptively across sessions. Reasoning uniquely predicted first session casual game scores for both groups and accounted for much of the relationship with WM. Furthermore, over time, WM became uniquely important for predicting casual game performance for the puzzle-based adaptive games but not for the speeded switching non-adaptive games. These results extend the burgeoning literature on cognitive abilities involved in video games by showing differential relationships of fluid abilities across different game types and extended play. More broadly, the current study illustrates the usefulness of using multiple cognitive measures in predicting performance, and provides potential directions for game-based cognitive training research.

Keywords: casual games; fluid intelligence; reasoning; skill acquisition; working memory.

Figure 1

General procedure for the entire casual game project (in blue) and the metrics for the current primary analysis (in red). Note, that the post-testing cognitive assessments were not used since the focus of the current study was on using baseline cognitive abilities to predict casual game performance across time.

Figure 2

Spearman correlation between game performance at each training session and each pre-training baseline measure. Error bars are 95% bootstrapped confidence intervals.

Figure 3

Multiple regression plots showing the predicted values for first (left) and final (right) session casual game performance for non-adaptive and adaptive group games derived from regression models using WM and reasoning as predictors. R² = adjusted R². The shaded area represents the 95% confidence region for each predictor in the model.