Narrative Review Connects Understanding to Chunking Complex Knowledge

A narrative review argues that understanding is not a separate mental faculty but an outcome of how the human cognitive system reorganizes complex knowledge. Within the established framework of cognitive load theory, understanding begins when learners combine multiple interacting elements into a smaller number of organized chunks that can be stored in long-term memory and later retrieved to guide action. The review integrates the concept of understanding into that existing architecture through working-memory limits, element interactivity and expertise, offering a mechanistic account of why some material can be understood while other difficult material may only be memorized.

Teaching students to learn with understanding is widely accepted as an important educational goal, yet the review notes that understanding has remained difficult to place within descriptions of human cognitive architecture. Earlier approaches describe it as an active process through which people construct meaning, with the resulting interpretation varying among learners. They distinguish deep-level processing, in which a learner attends to the intended meaning of a scientific problem or principle, from surface-level processing based on repetition, copying and other rote strategies. Some theoretical accounts treat intention as the key distinction, whereas others regard understanding and memorization as intertwined activities whose order can change as learners develop. The review treats these accounts as background to a more specific question: how can understanding be represented within a cognitive architecture built around limited working memory and extensive long-term memory?

The authors distinguish two broad categories of information. Biologically primary information, such as the ability to understand and use a native language, is acquired through ordinary social interaction and problem solving and is not normally dependent on formal instruction. Biologically secondary information, including reading, writing, mathematics and historical explanations, generally must be taught. The paper does not propose that every act of learning should be called understanding. Instead, it asks how the cognitive architecture of a learner can process highly complex secondary knowledge when the learner lacks the schemas needed to handle it efficiently. Novel secondary information must first pass through working memory, whereas information acquired through socially transmitted teaching can deliver knowledge that would otherwise require lengthy problem solving.

Within cognitive load theory, the central variable is element interactivity: the extent to which several elements must be considered simultaneously to make sense of the material. Solving an equation such as (a + b)/c = d for a requires the learner to coordinate the variables, their relationships and the rules of algebra. This makes the problem highly element-interactive. Learning vocabulary in a second language may be difficult because many separate words must be acquired, but each word can be learned independently. The authors therefore distinguish task difficulty from element interactivity. A large collection of independent components can be very difficult to learn, while a smaller set of components may be difficult because their relations must be held and manipulated together. The definition of element interactivity concerns the material itself and does not depend on the psychological consequences of processing it.

That distinction changes the proposed meaning of understanding. The term is applied to high-element-interactivity material, where learners must integrate multiple components, but not normally to low-interactivity material such as isolated vocabulary items. Someone who has forgotten that “le chat” means “the cat” has forgotten a fact, yet it would ordinarily be said that the person does not fail to “understand” that translation because no interacting relationships need to be constructed. By contrast, learning French grammar requires integration among many rules and structures, making understanding a relevant description. The framework thus links understanding to a specific information-processing achievement rather than attaching it to every successful act of recall. Difficulty alone does not establish that the learner is engaged in understanding.

Cognitive architecture explains why that achievement is demanding. When novel secondary information is encountered, it must first be processed in working memory, which has sharply limited capacity and duration. The paper cites estimates that working memory can hold about seven items but can simultaneously manipulate only about three or four. Novel, highly interactive material can therefore exceed a learner’s processing capacity, whereas low-interactivity information generally imposes a lighter working-memory load. Expertise changes the effective complexity of the same material. For a novice, an algebra problem contains several interacting elements; an experienced solver retrieves the organized procedure as a single chunk. Long-term memory, by contrast, has no known capacity or duration limits and can preserve extensive integrated knowledge, including solutions that were once difficult to understand.

The proposed mechanism of understanding is chunking. After information has been processed in working memory, it can be stored in long-term memory, where multiple smaller components may be combined into larger units. For the novice, those units remain separate and highly interactive; for the expert, they have become a familiar chunk that can be transferred back into working memory and used to guide action. On this account, understanding improves performance and lowers cognitive load because the information’s effective element interactivity has changed, rather than because an abstract feeling of comprehension has been added to the system. A stored chunk can be retrieved as a single element, or as a relatively small number of elements, allowing working memory to process material that previously required simultaneous coordination of many parts. The paper connects this account to generative learning models in which selected information is organized in working memory and integrated with knowledge already held in long-term memory.

The review also draws a practical implication from established cognitive load theory effects, most of which concern complex material that requires understanding. A worked example can reduce extraneous load for novices by showing a limited number of challenging steps without requiring them to search randomly for solutions. By contrast, asking novices without relevant knowledge to solve unfamiliar problems may divert limited working-memory resources toward trial-and-error, leaving fewer resources available for interacting with the material itself. The isolated elements effect provides a more striking example. In a study cited by the paper, learners initially presented with the fully interacting form of very complex information learned less and used it less effectively than learners who first rote-learned separate elements and subsequently combined them. On the authors’ interpretation, initial memorization supported later understanding rather than competing with it, because separately learned elements could eventually be combined into an organized whole.

The proposal remains a theoretical synthesis rather than a new experimental demonstration. The article is explicitly a narrative review with no associated dataset, and its instructional claims draw partly on previously conducted randomized controlled trials and the wider body of cognitive load theory research. It also distinguishes understanding from transfer. Successfully using knowledge in a new problem can provide evidence of understanding, but failure to recognize its relevance in another problem does not by itself show that understanding was absent; the new task may require additional knowledge not yet stored. A learner may fully understand how to solve one equation but fail to recognize that the procedure is one of the steps needed in another, more complex equation. The paper therefore treats understanding and transfer as related but distinct outcomes and calls for additional experimentation to determine when additional knowledge enables additional transfer.

Its central claim is that understanding is already embedded in the cognitive architecture used by cognitive load theory to explain learning. The account is deliberately narrower than a universal account of meaning: understanding applies when highly interactive, biologically secondary information is processed, organized into chunks and stored in long-term memory. Once consolidated, that information can be retrieved efficiently and used to govern action in a way that would exceed novice working-memory capacity. The review also identifies boundaries for instruction. Its strategies are most relevant when material has high element interactivity, learners lack relevant prior knowledge and understanding is an intended outcome; they are less critical for low-complexity material that can be learned independently. On this reading, teaching for understanding means designing instruction around the architecture that makes comprehension possible while allowing memorization to serve as a step in the construction of an integrated knowledge structure.

Subject of Research: Cognitive load theory of understanding, chunking, working memory and expertise

Article Title: The Role of “Understanding” in Human Cognitive Architecture: A Cognitive Load Theory Perspective

Article References: Chen, O., Marcus, N., & Sweller, J. (2026). The Role of “Understanding” in Human Cognitive Architecture: A Cognitive Load Theory Perspective. Educational Psychology Review, 38(1), 135. https://doi.org/10.1007/s10648-026-10233-5

Image Credits: AI Generated

DOI: 10.1007/s10648-026-10233-5

Keywords: cognitive load theory, understanding, chunking, working memory, long-term memory, element interactivity, expertise, learning science, instructional design, narrative review, Role, Human