Working Memory and Learning — What Research Tells Us

Written by TOMOY · Published: April 8, 2026 · Updated: April 8, 2026

1. Working Memory as a Gateway to Long-Term Memory

To learn something new, you must first hold it in working memory long enough to process it. WM is the active workspace where new information is connected to what you already know, manipulated, and encoded into long-term memory. Without sufficient working memory capacity to hold the elements of new material simultaneously, the connections that constitute understanding cannot form.

This is why WM is considered foundational to learning — not because it stores what is learned, but because it is the system through which learning happens. Reading comprehension, following a lecture, solving a novel math problem, learning a new procedure: all require holding intermediate steps and partial results in WM while continuing to process incoming information.

2. Cognitive Load Theory

Sweller (1988) proposed cognitive load theory to explain why some instructional designs produce learning more efficiently than others. The central insight is that when the total cognitive load placed on working memory exceeds its capacity, learning suffers — not because learners are unmotivated or inattentive, but because the cognitive architecture responsible for learning has been saturated.

Sweller distinguished three types of load:

• Intrinsic load: the inherent complexity of the material itself — how many interacting elements must be held in WM simultaneously. High-element interactivity material (e.g., solving a multi-step equation) imposes high intrinsic load.
• Extraneous load: load generated by the way material is presented, not by the material itself. Poor design — redundant information, split-attention layouts, unclear explanations — adds extraneous load that consumes WM without contributing to learning.
• Germane load: cognitive effort directed at processing and encoding the material into long-term memory. This is the load that produces learning; the goal of good instruction design is to maximize germane load by minimizing extraneous load.

3. Reading Comprehension

Reading requires holding the beginning of a sentence while parsing its end, tracking pronoun references across sentences, and maintaining the gist of earlier paragraphs while processing new ones. All of these place continuous demands on the phonological loop and central executive.

Sentences with complex syntactic structures — long embedded clauses, multiple negations, passive constructions — impose higher WM demands than simpler alternatives conveying the same information. This is one reason that clear, direct writing is not merely a stylistic preference: it reduces the WM load required for comprehension and leaves more capacity for engaging with the content itself.

4. Mental Arithmetic

Arithmetic performed mentally is a direct exercise in working memory. Carrying intermediate results while performing subsequent operations, tracking where you are in a multi-step calculation, and keeping partial sums in mind while computing new ones all require active phonological loop rehearsal and central executive coordination.

As arithmetic becomes automated through practice, it draws less on WM — which is part of why fluency with basic facts frees up capacity for higher-order mathematical reasoning. The capacity freed by automaticity can then be applied to the parts of the problem that still require deliberate processing.

5. Strategies That Reduce WM Load During Learning

Several well-documented strategies help manage cognitive load during learning by offloading work from working memory:

• Chunking: organizing individual items into meaningful units reduces the number of independent elements that must be held in WM simultaneously. Expertise is partly constituted by larger and more elaborate chunks in long-term memory.
• Worked examples: studying a completed solution reduces the need to generate solution paths from scratch, freeing WM to attend to the structure of the solution rather than searching for it.
• Externalizing: writing down intermediate steps, using diagrams, or keeping a written record of where you are in a process offloads those representations from WM to the environment.
• Spaced practice: revisiting material over time consolidates it into long-term memory, gradually reducing the WM demands required to access and use it.