Reaching for the Unreachable: Reorganization of Reaching with Walking

Abstract

Previous research suggests that reaching and walking behaviors may be linked developmentally as reaching changes at the onset of walking. Here we report new evidence on an apparent loss of the distinction between the reachable and nonreachable distances as children start walking. The experiment compared nonwalkers, walkers with help, and independent walkers in a reaching task to targets at varying distances. Reaching attempts, contact, leaning, and communication behaviors were recorded. Most of the children reached for the unreachable objects the first time it was presented. Nonwalkers, however, reached less on the subsequent trials showing clear adjustment of their reaching decisions with the failures. On the contrary, walkers consistently attempted reaches to targets at unreachable distances. We suggest that these reaching errors may result from inappropriate integration of reaching and locomotor actions, attention control and near/far visual space. We propose a reward-mediated model implemented on a NAO humanoid robot that replicates the main results from our study showing an increase in reaching attempts to nonreachable distances after the onset of walking.

Keywords: Infant reaching; near and far space integration; perceived reachability; reaching and walking.

Fig. 1

Superimposed images from two cameras showing the experimental setup for the behavioral study. The first camera (large image) was located directly above the theatre, and recorded a birds-eye view of infants’ reaching movements and provided a clear view of the moment of contact. The second camera (small image in the right corner) was placed on the side, captured the side view of the child, and was used to determine the infants’ leaning angle.

Fig. 2

Mean proportion of trials in which 12-month-olds reached only, reached and leaned, or neither reached nor leaned. Leaning without reaching virtually never occurred and so is not shown. (a) Nonwalkers. (b) Walkers with help. (c) Independent walkers.

Fig. 3

Attempted boundary, that is the longest distance at infants attempted to reach on .50 or more of the trials and the Contact boundary, that is the farthest distance at which contacts are made on .50 or more of the trials for infants. (a) The Attempted boundary. (b) The Contact boundary.

Fig. 4

General scheme of our reward-based learning architecture for near and far space representation. Two separate three-layer neural networks are used to approximate the state-action mapping function. The state here is depth-cue information, whereas action is equivalent to a specific depth estimate that can be associated with reaching and/or walking action.

Fig. 5

Examples of learned Q-value functions for near space representation network. A colormap shows reward predictions (Q-values) of the output neurons (x-axis, n = 18 neurons) for all distances used during training (y-axis). Figure on the right shows how the representation of near space has changed after the onset of walking; the output neurons that previously encoded the distances near the boundary of reachable space, activate now also for distances outside of the reachable space. (a) Before walking (b) After walking.

Fig. 6

Percent of reaches with standard deviation (range of error bars) for infants with nonwalkers (on the left) and independent walkers (on the right) to 5 different distances (30, 37, 47, 60, and 70 cm). The computational model was tested with the use of the NAO humanoid robot, and the results are provided here for comparison. Since the NAO robot is much smaller than an average 12-month-old infant, testing distances were adjusted accordingly (13, 15, 21, 27, and 29 cm). Additionally, we tested the robot with two far distances (37 and 39 cm). (a) Nonwalkers. (b) Walkers.