6.1 Models of Working Memory

Working memory is the mental workspace you use to hold and manipulate information in real time. Without it, you couldn't follow a conversation, solve a math problem in your head, or even read this sentence while remembering how it started. Two major models attempt to explain how this system works, and they take surprisingly different approaches.

Components and Functions of Working Memory Models

Components of Baddeley's working memory model

Baddeley's model breaks working memory into four distinct components, each handling a different job.

• Central executive is the attentional control system. It coordinates the other components, allocates cognitive resources, and decides what gets priority. Think of it as the supervisor that directs traffic between the subsystems below.
• Phonological loop stores and rehearses verbal and acoustic information. It has two parts: a phonological store that holds sound-based information, and an articulatory rehearsal process that refreshes it (like silently repeating a phone number to yourself).
• Visuospatial sketchpad maintains and manipulates visual and spatial information. It also has subcomponents: one for visual features (what things look like) and one for spatial relationships (where things are and how they move).
• Episodic buffer was added to the model in 2000 to solve a problem the original three components couldn't handle. It integrates information across the other subsystems and links them to long-term memory. It has its own limited storage capacity and is controlled by the central executive.

Functions of the phonological loop and sketchpad

Phonological Loop:

The phonological store holds speech-based information for roughly 1–2 seconds before it decays. To keep information alive past that window, the articulatory rehearsal process "replays" it internally. This rehearsal process also converts written words into a phonological code, which is why you might "hear" words in your head while reading.

Two classic effects demonstrate how the loop works:

• Word length effect: You can remember fewer long words than short words because longer words take more time to rehearse, and the store decays before you finish the loop. Try remembering a list containing "university, hippopotamus, refrigerator" versus "cat, pen, dog."
• Phonological similarity effect: Words that sound alike (e.g., cat, hat, mat, bat) are harder to keep distinct in the phonological store, leading to more recall errors. Words that look similar but sound different don't cause the same problem.

Visuospatial Sketchpad:

This subsystem splits into two parts. The visual cache stores information about what things look like (color, shape, texture), while the inner scribe handles spatial information and movement sequences, like mentally tracing a route through a building.

The sketchpad holds roughly 3–4 objects at a time. Its capacity is selectively disrupted by visual or spatial interference tasks. For example, if you're holding a mental image of a map and someone asks you to judge the orientation of a rotated shape, performance drops because both tasks compete for the same sketchpad resources. A verbal task, by contrast, wouldn't interfere much because it draws on the phonological loop instead.

Central executive in working memory

The central executive is the most important and least understood component of Baddeley's model. It doesn't store information itself. Instead, it controls how attention and resources are distributed across the other subsystems.

Its key functions include:

• Focusing attention on relevant information while inhibiting distractions
• Switching attention between tasks or between subsystems
• Dividing attention when you need to handle two things at once
• Coordinating the subsystems, deciding when to route information to the phonological loop versus the sketchpad
• Retrieving information strategically from long-term memory when needed

Because the central executive has limited capacity, it becomes a bottleneck under high cognitive load. This is why multitasking degrades performance: the executive can only manage so many demands before errors increase and processing slows down.

Models of working memory

The two dominant models take fundamentally different approaches to explaining the same phenomenon.

Baddeley's Multicomponent Model is a structural model. It proposes that working memory consists of separate, specialized storage systems (the loop, the sketchpad, the buffer) coordinated by a central executive. Each subsystem has its own capacity limit and handles a specific type of information.

Cowan's Embedded-Processes Model is a functional model. Rather than proposing separate storage bins, Cowan argues that working memory is simply the activated portion of long-term memory. Within that activated region, a smaller focus of attention holds about 3–4 chunks of information at any given time. There are no dedicated subsystems for verbal versus visual material; the limit is a general attentional one.