Digit-Tracking Reveals Curiosity-Driven Visual Attention in Macaque Monkeys

Abstract

We used digit-tracking, a touch-based method for assessing visual attention, to investigate spontaneous exploration in macaque monkeys. By engaging with degraded images on a touch-sensitive display, monkeys could uncover high-resolution portions through finger movements, allowing for natural and unrestricted interaction. Monkeys received juice rewards after touching a predetermined number of pixels, but no specific regions were targeted. Attention maps were generated from their interactions, along with data from human digit-tracking and monkey eye-tracking experiments. We applied a saliency model and a Convolutional Neural Network (CNN) model to predict the empirical explorations. The correlation between model prediction maps and empirical attention maps indicated that monkeys focused non-randomly on information-rich regions, with the CNN model providing the most accurate predictions. These findings suggest that exploration was driven by intrinsic curiosity, beyond the extrinsic rewards for interaction. Digit-tracking offers a minimally invasive, portable alternative to eye-tracking, expanding research opportunities in visual cognition within ecologically valid settings.