Posture affects how robots and infants map words to objects

Area of Science:

• Cognitive Science
• Developmental Psychology
• Robotics

Background:

• Infant word learning involves mapping words to referents and remembering these associations.
• Spatial location is increasingly recognized as crucial for infants' word-learning processes.
• Existing models often separate spatial information from word-object binding.

Purpose of the Study:

• To propose and test a new theoretical model of infant word learning centered on body-centric spatial location.
• To investigate the role of momentary body posture in binding heard names with visual objects.
• To determine if body-centric spatial contingency influences name-object mapping over temporal contingency in infants.

Main Methods:

• Utilized a name-object mapping task with human infants and a humanoid robot model.
• Tested scenarios involving absent targets and targets in previously associated foil locations.
• Employed a humanoid robot to instantiate and test the body-centric spatial hypothesis, comparing its performance to infant data.

Main Results:

• The humanoid robot model replicated existing infant data and generated novel predictions.
• Infants prioritized body-centric spatial contingency over temporal contingency for name-object mapping, even when space was task-irrelevant.
• Both infants and the robot demonstrated memory for name-object mappings in new spatial locations.

Conclusions:

• Body-centric spatial location and momentary posture are central to infant word learning and memory.
• The proposed model explains how memory emerges through the body's disposition in space, not by separating it from the word-object mapping.
• This research offers a new perspective on how embodied cognition facilitates early language acquisition.