Many cognitive assessments are limited by their reliance onrelatively sparse measures of performance, like per-item ac-curacy or reaction time. Capturing more detailed behavioralmeasurements from cognitive assessments will enhance theirutility in many settings, from individual clinical evaluationsto large-scale research studies. We demonstrate the feasibilityof combining scene and gaze cameras with supervised learn-ing algorithms to automatically measure key behaviors on theblock design test, a widely used test of visuospatial cognitiveability. We also discuss how this block-design measurementsystem could enhance the assessment of many critical cogni-tive and meta-cognitive functions such as attention, planning,progress monitoring, and strategy selection.