Potential of cross-task stylus-based drawing features for predicting psychological vulnerability in children

The present study examined whether behavioral features extracted from a digital Mandala coloring task could predict children’s stress-related vulnerability group status (stress-alert vs. non-alert), as independently assessed through the PITR (Person-in-the-Rain) drawing task. Using both MVPA (Multivariate Pattern Analysis) and mixed-effects logistic regression, we found a weak but statistically significant, above-chance cross-task signal distinguishing the stress-alert group from the non-alert group, although the effect sizes were small. Specifically, the logistic regression identified mean stroke length as the only significant predictor, and in repeated cross-validation a linear-kernel MVPA classifier achieved a mean balanced accuracy modestly above chance. These results indicate a population-level rather than an individually diagnostic signal, suggesting that subtle motoric patterns in emotionally neutral drawing tasks may partly reflect underlying psychological states. We discuss the theoretical, clinical, and methodological implications for the consistency of cross-task emotional signals, non-invasive assessment, and early screening in child populations.