The Mind's Eye: Mental Models in Cognition and Consciousness

Source: Yoonnam / AdobeStock

This is Part 2 of a five-part blog series exploring the nature of cognition and its relationship with consciousness. In Part 1, we considered the surprisingly elusive definition of cognition as well as the relations between cognition, emotion, intelligence, and (in a preliminary way) consciousness. In Part 2, we will consider a newer theory by neuroscientist Joseph LeDoux that attempts to provide a more specific definition of cognition, one that emphasizes the role of mental models, recognizes the importance of unconscious mental processes, and can be used to construct a theory of consciousness.

Internal (mental) models as the basis for cognition

Joseph LeDoux has written extensively on the evolution of cognition. He defines cognition as “the ability to use internal representations of information to construct mental models of the world.” In other words, cognition entails “the capacity to construct models of the world and use these in thinking, planning, deciding, acting, and even feeling.”¹

Internal representations—elements of mental models—are like maps or images of the body and the world, constructed by the brain from sensory input. Of course, these are not literal pictures in the brain but distributed patterns of neural activity, often involving coordinated interactions among brain regions that may be physically distant from one another. Mental models are formed by the assembly of such representations into context-specific simulations that integrate perceptions, memories, and bodily states. These models allow animals to mentally rehearse possible outcomes and thus select actions based on anticipated consequences—a major leap beyond purely reactive behavior. For example, a squirrel hoarding nuts for the winter may simulate where to hide them and how to retrieve them later.

In LeDoux’s view, an organism qualifies as cognitive only if it can construct and use such internal models. By this standard, simple information processing, learning, or even complex instinctual behavior does not suffice. Other scientists set the bar lower, counting as “cognitive” any organism capable of information processing or learning—including protostome invertebrates (e.g., worms, insects, mollusks) and even single-celled organisms. LeDoux disagrees: “Simply having information processing and learning capacities is not enough for membership in the cognitive club. Nor is the ability to exhibit complex behavior.”^2,3 He notes that even habitual and instinctual behavior can appear quite complex and is not in itself evidence of true cognition.

Model-free precursors of model-based cognition

Before model-based cognition evolved, evolution had already established model-free learning mechanisms in many species.

• Model-free learning generates habits by simply linking an action to its past outcome, so cues trigger the same response in similar conditions.
• Model-based learning, in contrast, constructs an internal model from stored information and uses it to evaluate options and guide goal-directed behavior.

Model-based predictions support decisions made in the moment, rather than through habits alone, giving organisms far greater flexibility. If goals change or a behavior stops working, the brain can explore alternatives within its simulation model and select a new course of action. Behaviors learned in one context—such as searching for food—can also be adapted to very different needs, such as navigating new terrain or evading predators. In this process—which is a big advance from model-free learning—working memory systems temporarily store and manipulate task-relevant information, while executive control⁴ selects and maintains the memories most likely to generate successful options.

For LeDoux, the ability to guide behavior with internal models marks the true evolutionary emergence of cognition.

The evolution of model-based cognition

LeDoux notes that it is well-established that mammals, in pursuing goals, use internal representations to construct mental models simulating the outcomes of future actions⁵ and there is evidence that some kinds of birds do, too,⁶ whereas he thinks it is likely (though not completely certain) that all lower vertebrates and protostome invertebrates are limited to model-free reinforcement learning and habitual behavioral control. It is not known when mental modeling emerged in mammals, but once it appeared, it likely elaborated further as different mammalian lineages faced new ecological challenges and evolved different body plans.

Mammals and birds appear to have evolved model-based cognition independently, each in parallel, by modifying precursors inherited from their respective vertebrate ancestors.⁷ Those precursors would have been model-free mechanisms—especially the mechanisms of associative learning and habit formation. This parallel evolution in mammals and birds may have been driven by the high metabolic demands of being warm-blooded, which would have created selective pressure for more efficient foraging strategies.⁸

Refining Kahneman’s Thinking, Fast and Slow into a three-level model

Modern theories of cognition often distinguish between two types of mental processes: intuition and deliberation. Intuition is fast, effortless, automatic, and happens without conscious awareness, while deliberation is slower, effortful, and conscious. This distinction was popularized by Daniel Kahneman in his 2011 book Thinking, Fast and Slow, which describes these as two different systems for processing information. Kahneman’s System 1 (fast) handles intuitive judgments, routine decisions, and pattern recognition. System 2 (slow) is used for reasoning, problem-solving, and self-control. System 1 quickly generates impressions and feelings, while System 2 can evaluate and, if needed, override these. Both systems are forms of cognition according to Kahneman, but they differ in speed, effort, and the level of conscious control involved.

Cognition Essential Reads

Daily Prophets: How Your Brain Predicts the Future

The Upside of Rumination

Kahneman considered both Systems 1 and 2 cognitive. LeDoux, however, relocates certain fast processes—like intuition and non-conscious reasoning—from Kahneman’s System 1 into a new System 2 of unconscious cognition, reserving System 1 for non-cognitive habits and System 3 for conscious thought.⁹

Kahneman vs. LeDoux systems comparison table

Source: adapted by Lewis from Kahneman (2011) and LeDoux (2023)

From cognition to consciousness

In LeDoux’s theory, which he integrates with higher-order theories of consciousness, System 1 constitutes first-order non-cognitive processes within what he calls the neurobiological realm. System 2 involves higher-order but pre-conscious cognitive processes within the cognitive realm. System 3 consists of higher-order conscious states within the conscious realm. Note that both Systems 2 and 3 involve goal-directed behavior controlled by higher-order cortical systems, whereas System 1 involves innate/reflexive/habitual behavior and is controlled mostly via subcortical brain areas.

Yet, additional ingredients are still needed for cognition to cross the threshold into consciousness. While the exact mechanisms of consciousness remain unknown, LeDoux makes headway toward a plausible account by identifying some of those ingredients. We will explore his theory of consciousness in Part 3.