KembaraXtra – Psychology: Computational Theory

Definition

A computational theory is a formal account of how information is transformed in order to solve a psychological or perceptual problem. In vision science, the term is strongly associated with David Marr’s theory of visual processing, which sought to explain how patterns of light reaching the retinas are converted into internal representations of objects, surfaces, depth, colour, and movement.

Structure and Mechanism

Marr distinguished among three levels of explanation: the computational level, which specifies the problem being solved; the algorithmic level, which describes the representations and procedures used; and the implementational level, which concerns how the process is physically realized in the nervous system. His visual theory proposed a progression from a primal sketch of basic image features, through a viewer-centred two-and-a-half-dimensional sketch, to a three-dimensional representation of object structure.

Role in Cognitive Psychology and Neuroscience

Computational theories provide precise models that can generate testable predictions about perception, memory, language, reasoning, and action. They encourage researchers to separate the goal of a cognitive system from the specific algorithm and neural machinery that achieve it. This distinction has strongly influenced cognitive science, artificial intelligence, and computational neuroscience.

Research Significance and Limitations

Formal models can clarify assumptions and reveal whether a proposed mechanism is sufficient to produce observed behaviour. However, computational theories may oversimplify biological systems or rely on representations that are difficult to verify directly. Strong explanations therefore integrate computational analysis with behavioural findings, neural evidence, and environmental context.

Summary

A computational theory explains cognition or perception as a structured transformation of information. Marr’s theory of vision is a classic example, distinguishing computational, algorithmic, and implementational levels. Such theories support precise modelling and testable predictions. Their value is greatest when linked to behavioural and neurological evidence.