Towards a “Treadmill Test” for Cognition: Reliable Prediction of Intelligence From Whole-Brain Task Activation Patterns

Abstract

Identifying brain-based markers of general cognitive ability, i.e., “intelligence”, has been a longstanding goal of cognitive and clinical neuroscience. Previous studies focused on relatively static, enduring features such as gray matter volume and white matter structure. In this report, we investigate prediction of intelligence based on task activation patterns during the N-back working memory task as well as six other tasks in the Human Connectome Project dataset, encompassing 19 task contrasts. We find that whole brain task activation patterns are a highly effective basis for prediction of intelligence, achieving a 0.68 correlation with intelligence scores in an independent sample, which exceeds results reported from other modalities. Additionally, we show that tasks that tap executive processing and that are more cognitively demanding are particularly effective for intelligence prediction. These results suggest a picture analogous to treadmill testing for cardiac function: Placing the brain in an activated task state improves brain-based prediction of intelligence.