What Is a Psychological Theory?

Kurt Lewin said that there is nothing so practical as a good theory. Psychological theories are valuable for guiding practice in education, mental health, business, and other domains. They provide answers to intrinsically interesting questions concerning many kinds of thinking including perception, emotion, learning, and problem-solving.

A recent article by Kurt Gray offers to help psychologists develop better theories by means of theory maps that display associations between different factors. For example, the theory map for moral judgment identifies a positive association between feeling empathy and seeing a vulnerable patient, and a negative association between feeling empathy and emotion regulation.

Such associations may be empirically valid, but associations do not provide the causal information that serves to explain why people do what they do, and that guides interventions in the world to deal with practical problems. Maps of association fall short of explanatory theories.

So what is a psychological theory? In physics, theories are usually sets of mathematical equations that identify causal factors. For example, Newton's theory of planetary motion states equations for characterizing how forces such as gravity make planets move. Psychology is too messy to admit theories of this kind.

Better methodological role models come from biology and medicine. Theories in biology usually consist of descriptions of mechanisms, which are combinations of connected parts whose interactions produce regular changes. For example, in genetics, the parts are cells, genes, proteins, and other entities that interact to produce inheritance and selection. Medical theories are aimed at explaining and treating disease and usually operate by indicating how mechanisms that are useful for bodily functioning can break down, for example when arteries are blocked during a heart attack.

Analogously, we can look for mechanistic explanations in psychology at two levels: representational and neural. Since the cognitive revolution in the 1950s and 1960s, psychology has employed explanations in terms of mental representations such as concepts, schemas, and rules. These are the parts of mental systems that interact with each other through computational processes such as inference and spreading activation. The operations of these psychological mechanisms have been spelled out with sufficient precision to run computer simulations on cognitive architectures such as ACT and Soar.

A cognitive architecture is a general proposal about the representations and processes that produce thought. A good strategy for psychologists who want to theorize about their empirical work is to try to show how phenomena can be causally explained by the operations of a cognitive architecture. But not all cognitive architectures invoke symbolic, verbal thinking.

Since the 1980s, psychology has increasingly developed explanations in terms of neural mechanisms, where the parts are neurons, the connections are synapses, and the interactions include excitation and inhibition that lead to patterns of neural firing. Prominent neural network cognitive architectures include PDP (parallel distributed processing), Leabra, and CLARION. Hence an alternative strategy for psychological theorizing is to show that phenomena result from neural mechanisms.

My favorite cognitive architecture is Chris Eliasmith’s Semantic Pointer Architecture, which combines rich neural mechanisms with the ability to model high-level representations such as concepts and symbolic inferences such as deduction. It has many explanatory applications, ranging from motor control to consciousness.

All of these cognitive architectures are genuine theories in that they provide descriptions of mechanisms that causally explain a wide range of phenomena. These mechanisms say why things happen and provide guidance on how to intervene in the world to bring about practical benefits. So, they are more effective both theoretically and practically than sets of associations. Psychology does indeed need better theories, but they should be ones that specify causal mechanisms.