Dynamics of infant preshaping: Longitudinal analysis using automated hand pose tracking from home video

Infant reaching and grasping scaffold opportunities for exploration, learning, and communication, making them central to early development. Yet most prehension research relies on laboratory methods, which constrain sample sizes and the density of longitudinal follow-up. Here we used automated pose estimation (HaMeR) to extract markerless hand-pose features from in-home videos recorded by caregivers biweekly from 6 to 13.5 months of age (N = 63 infants; 881 videos; M = 13.98 videos per infant), tracking hand configuration across normalized reach time to reveal the within-reach dynamics of preshaping. We examined two components of preshaping: the opening between the thumb and index finger (aperture) and the tucking of the supporting fingers (supporting digit flexion). Flexible nonlinear modeling revealed distinct developmental trajectories for each component. By 6 months, infants were already tucking the supporting fingers about halfway through the reach, and by 11 months this adjustment shifted earlier. Anticipatory aperture control, in contrast, did not emerge until around 9-10 months, after which the moment of peak opening also shifted progressively earlier in the reach. The two components thus followed distinct developmental timelines, indicating partially independent control processes and suggesting that supporting digit flexion consolidates earlier, providing a foundation for the later refinement of grasping. Pairing in-home assessment with machine-learning-based pose analysis reveals developmental dynamics that laboratory methods cannot capture at this density, opening new opportunities to study infant motor development at scale in naturalistic settings.