|dc.description.abstract||Developmental researchers typically use equivocal terminology to explain the foundation of visual perception in infancy. Terms such as “occlusion”, “permanence”, and even “object” provide a convenient explanation of infant behaviour to a learned audience. However, they also afford the infant unstated, and perhaps undeserved knowledge of the world. Are infants born with the ability to individuate and identify mid-sized objects? How is an object distinguished from its surroundings? This thesis seeks to explore potential explanations for infant behaviours without using ambiguously leading terminology, and in so doing, understand the explanatory power of these post-hoc linguistic definitions of object knowledge.
This thesis investigates some of these questions through an interpretive approach that is applied to behavioural experiments involving the development of visual perception in infancy. These experiments typically consist of a procedure in which object knowledge is attributed to infants based on preferential looking towards one of two different displays. What can be obscured or ignored by the interpretations of developmental researchers is the possibility that object knowledge has developed from lower level primitives. This thesis attempts to analyse object-related concepts and describe them “computationally” using primitives such as position and trajectory.
A bottom-up approach to interpreting visual perception is developed where a concept is described by its observed input and output variables, and a plausible mechanism for the conversion from the former to the latter is deduced. Sensory input consists of low level positional change and simple motor action. Higher level mechanisms are then built on top of this foundation. The bottom-up interpretive approach was developed iteratively during the course of this research, and is considered a novel contribution of the thesis.
The bottom-up interpretation of multiple experiments over several iterations of the research process resulted in a preliminary theoretical model of early visual perception. The model consists of perceptual primitives that explain commonly assumed object-related principles including continuity, cohesion, and segregation. Importantly, it does so using a foundation consisting only of visible positional change and motor action so that a post-hoc linguistic definition can provide no additional explanatory power.||en_NZ