'Samuelson et al. (2011), continuing the work of Baldwin (1993), showed that there are clear posture and spatial biases in infants' learning of the mapping between words and objects.
In their experimental setup, infants repeatedly experience two new objects (the target and the foil) in consistent but different locations. Subsequently they hear the object name 'modi' while attending to the foil object which has now been placed in the location normally associated with the target object. On testing with both objects present in new locations, the infant is asked 'where is the modi?'. The statistically significant majority of children select the target object.
This means that they are selecting the object normally associated with the spatial location they were attending to, rather than the actual object they were looking at when they heard the name. By implication, this means that infants cannot be using a simple mapping between the object features observed and the word detected at that point in time. They rather rely on a memory for own posture and the related object location to associate objects and their names.'
I can't help but picture George Dickie and Arthur Danto in diapers, sucking on a pacifier and calling every object in a certain place 'art', simply because that is the location that they learned to associate with the word.