Chapter 5: STM & Working Memory

The Modal Model of Memory (Atkinson & Shiffrin)

The modal model of memory says memory has three main structural parts and mental actions (control processes) that move information between them.

Two key terms you must separate:

• Structural features = the memory stores themselves

• Control processes = things you actively do with your mind

2. Structural features = the “boxes” in the model

These are called structural features because they are the basic architecture of memory.

The three structural features:

1. Sensory memory

   • Holds all incoming sensory information

   • Very brief (milliseconds–seconds)

   • You are not consciously aware of most of it

2. Short-term memory (STM)

   • Holds a small amount of information

   • Lasts ~10–15 seconds without rehearsal

   • What you are actively thinking about

3. Long-term memory (LTM)

   • Large capacity

   • Long-lasting (minutes → lifetime)

👉 These stores exist whether or not you choose to use them.

3. Control processes = what YOU do with memory

Control processes are flexible strategies you use to manage information.

They:

• Are not automatic

• Differ depending on the task

• Are under your control

Examples of control processes:

• Selective attention

• Rehearsal

• Chunking

• Using mnemonics

• Focusing on meaningful or important info

Think of control processes as mental tools you apply to the memory system.

4. Rehearsal

Rehearsal is a control process that operates on short-term memory.

Example:

• Repeating a phone number over and over so it doesn’t fade

Rehearsal:

• Keeps information active in STM

• Increases the chance it will move into LTM

5. Walking through the Rachel & pizza example (step by step)

This is where everything comes together.

🟦 Step 1: Sensory memory (Figure 5.3a)

Rachel looks at the webpage.

• All visual information enters sensory memory

• Includes ads, text, pictures, links

• Most of it will disappear almost immediately

She hasn’t chosen anything yet.

🟦 Step 2: Selective attention → Short-term memory (Figure 5.3b)

Rachel uses selective attention (a control process) to focus on:

👉 Mineo’s Pizza phone number

Because she attends to it:

• That number enters short-term memory

• The rest of the page fades away

This shows:

🟦 Step 3: Rehearsal keeps it in STM (Figure 5.3c)

Rachel repeats the number in her head:

This is rehearsal, another control process.

Rehearsal:

• Prevents the number from fading

• Keeps it available long enough to dial

• May help store it in long-term memory

6. Why this distinction matters (conceptually)

Structural features:

• Are passive

• Just hold information

Control processes:

• Are active

• Determine:

  • What gets noticed

  • What gets remembered

  • What gets ignored

This explains why two people can experience the same thing but remember different details.

Summary

This passage introduces three key memory processes:

1. Encoding → getting information into long-term memory

2. Storage → keeping information in long-term memory

3. Retrieval → bringing information out of long-term memory

You’ve already seen attention and rehearsal. These are just the next steps.

2. Step-by-step using Rachel’s example (very important)

🟦 Step 1: Rachel wants to remember the number later

This means:

• Holding it briefly in STM is not enough

• She needs it to last days, not seconds

So she decides to memorize it.

🟦 Step 2: Encoding → moving STM → LTM

This process is called encoding.

Encoding =

Important:

• Encoding is a control process

• It often involves strategies (rehearsal, meaning, organization)

• Just seeing something does not guarantee encoding

📌 Example:

• Repeating the number

• Grouping digits

• Linking it to something meaningful

🟦 Step 3: Storage → the number stays in LTM

Storage =

Once stored:

• The number can last days, months, or years

• It does NOT stay in awareness constantly

🟦 Step 4: Retrieval → remembering later

This is retrieval.

Retrieval =

Why STM?

• Because conscious awareness happens in STM

• You can’t think about something unless it’s in STM

📌 That’s why the book says:

3. Why STM and LTM must interact (core idea)

This paragraph is making a big theoretical point:

The flow looks like this:

1. Sensory memory → STM (via attention)

2. STM → LTM (via encoding)

3. LTM → STM (via retrieval)

You need STM:

• To encode information

• To retrieve and use information

4. Why forgetting phone numbers is so easy

The book reminds you:

That’s why:

• If you don’t rehearse → it fades

• If you don’t encode → it never reaches LTM

Rachel avoids forgetting by intentionally encoding the number.

Sensory Memory

Part 1: Sparkler + movie projector (intuition builder)

🔥 Sparkler trail

When you wave a sparkler:

• You see a continuous trail

• But physically, there is no trail in the air

What’s happening?
👉 Your brain briefly holds onto the image of the sparkler after it moves.

This is called persistence of vision:

⏱ Duration: fractions of a second

🎥 Movie projector

Movies are NOT smooth motion.
What’s really happening:

• Still image

• Darkness

• Still image

• Darkness
  (24 times per second)

You don’t see the darkness because:
👉 Your visual system fills the gaps

Again: persistence of vision

Key takeaway so far

Your visual system:

• Temporarily stores visual information

• Even after the stimulus disappears

• But very briefly

This leads to the next question…

Part 2: Sperling’s question (the real experiment)

George Sperling (1960) asked:

The setup (important to visualize)

He flashed this for 50 milliseconds (very fast):

X A C M
F D L N
Z T B P

12 letters total.

Whole report method (first test)

Task:
“Report as many letters as you can.”

Result:
👉 People reported about 4–5 letters

At first glance, this looks like:

But participants said something crucial…

What participants reported

They said:

This suggests:

• The problem is not seeing

• The problem is rapid fading

Part 3: Partial report (the breakthrough)

Sperling changed the task.

What changed?

• Letters flash

• Immediately after, a tone plays:

  • High tone → report top row

  • Medium → middle row

  • Low → bottom row

Participants don’t know in advance which row they’ll report.

🔔 Why this matters

They must:

• Briefly store ALL rows

• Then use the tone to select one

Partial report results

Participants correctly reported:

• ~3.3 out of 4 letters in the cued row

• That’s 82% accuracy

And this worked for any row.

🔑 Huge conclusion

If they can report 82% of any row, then right after the flash they must have had access to:

They just can’t hold them long enough to say them all.

Part 4: Delayed partial report (how long does it last?)

Now Sperling delayed the tone.

Delay = 1 second

Result:

• People could report only ~1 letter

📉 That means the visual information:

• decays rapidly

• Mostly gone within 1 second

Final conclusion: Sensory memory

Sperling concluded that we have:

🧠 Iconic memory (visual sensory memory)

• Holds most or all visual input

• Duration: less than 1 second

• Capacity: very large

• Extremely fragile

This is the sensory memory stage in the Atkinson & Shiffrin model.

Auditory version (echoic memory)

Sounds also persist:

• You say “What?”

• Then suddenly understand before it’s repeated

That’s echoic memory:

• Auditory sensory memory

• Lasts a few seconds

Why this matters for memory theory

This section proves:

1. Sensory memory is high-capacity

2. Sensory memory is very short-lived

3. Attention is required to move info into STM

4. Most information is lost unless selected

This sets up:
👉 Short-Term Memory (next section)

Exam summary (very important)

Short Term Memory: Storage

Two core questions:

1. How long does STM last?

2. How much can STM hold?

1⃣ What is the duration of short-term memory?

❌ Common misconception

People often say:

⚠ That is wrong in cognitive psychology.

✅ Actual definition

The key experiment: Peterson & Peterson (1959)

What they wanted to test

The procedure (step-by-step)

1. Participants see three random letters
   Example: F Z L

2. Immediately after, they see a number
   Example: 403

3. They must:

   • Count backwards by 3s
     (403, 400, 397, 394…)

👉 This prevents rehearsal
(no repeating “FZL, FZL, FZL”)

4. After a delay (3–18 seconds), they are asked:

The results (this is the key)

📉 Massive drop-off

What this proves

• Information rapidly disappears from STM

• Without rehearsal, STM:

  • Fades in 15–20 seconds

  • Is not durable

💡 This is why rehearsal is called a control process — it keeps STM alive.

Exam-ready conclusion

2⃣ How many items can STM hold?

Now we shift from time → capacity

Digit span (classic measure)

What is digit span?

Example:

• 7–4–1–9–2–6–8

Average digit span

• 5–9 items

• Roughly the length of a phone number

Miller (1956): “The Magical Number 7 ± 2”

George Miller summarized earlier research and proposed:

So:

• Minimum ≈ 5

• Maximum ≈ 9

This became the traditional STM capacity estimate.

More recent view: STM holds ~4 items

Later researchers argued:

Change detection experiments (Luck & Vogel, 1997)

Instead of digits or words, participants:

• Saw visual objects

• After a brief delay, detected whether something changed

Findings:

• Performance plateaus around 4 items

• Adding more items doesn’t improve accuracy

Cowan (2001) conclusion

The larger “7 ± 2” number likely reflects:

• Chunking

• Familiarity

• Long-term memory support

How these findings fit together

Summary sentence

The answer is chunking. I’ll explain this in a clean, logical flow and tie each experiment to the point it’s making.

1⃣ What Luck & Vogel showed (capacity is small)

The key result

• When participants had 1–3 items, performance was almost perfect

• When there were 4 or more items, accuracy dropped sharply

👉 Conclusion:

This supports Cowan’s estimate and challenges the older “7 ± 2” idea as a pure capacity limit.

So STM is:

• Small

• Strictly limited

2⃣ The big question this raises

If STM is so limited…

We clearly do it all the time — so something must be helping STM.

3⃣ Miller’s solution: Chunking

What is chunking?

Chunking is:

A chunk is:

So STM doesn’t magically get bigger —
it just holds fewer, larger units.

4⃣ Simple example: words → chunks

You’re given these 8 words:

At first:

• That’s 8 separate items → hard to remember

But if you reorganize them:

• ringtail monkey

• jumped wildly

• young child

• city zoo

Now you have 4 chunks instead of 8 items.

And if you go one step further:

Now it’s essentially 1 meaningful chunk.

5⃣ Why meaning matters

Chunking works because:

• Meaning creates strong associations

• Strong associations let elements be treated as one unit

This is why:

• Random words → STM ≈ 5–8 items

• Meaningful sentences → STM ≈ 20+ words

The number of chunks stays small — the content inside each chunk grows.

6⃣ Letters example (organizations)

A long string of letters is hard to remember…

Until you recognize it as:

• FBI

• CIA

• NASA

• WHO

Now:

• Same letters

• Fewer chunks

• Much easier to hold in STM

This is chunking based on long-term knowledge.

7⃣ Ericsson et al. (1980): the extreme case

Participant: S.F.

• Normal digit span: 7 digits

• After extensive training: up to 79 digits

🚨 This did NOT mean his STM capacity increased.

Instead:

• He re-coded digits into meaningful chunks

• Used long-term memory knowledge (e.g., running times)

Examples:

• 3492 → “3 minutes 49.2 seconds” (mile time)

• 893 → “89.3, very old man”

This shows:

8⃣ The critical theoretical point

Chunking shows an interaction between STM and LTM:

• STM holds the chunks

• LTM provides meaning and structure

• Together, they allow complex cognition

STM alone is limited — but STM + LTM is powerful.

9⃣ Why this matters in everyday life

You use chunking constantly:

• Reading letters as words

• Words as phrases

• Phrases as ideas

• Phone numbers as area codes + exchanges

• Conversations as topics

Without chunking, thinking would grind to a halt.

🔑 One-sentence exam summary

1⃣ The core issue: items vs. information

Traditional view

STM capacity is described as:

• ~4 items (Luck & Vogel; Cowan)

• or 5–9 items (Miller)

This treats all items as if they are equal in size.

New proposal

Some researchers argue that STM capacity should instead be described as:

In visual STM, information = visual features/details, such as:

• Color

• Shape

• Orientation

• Shading

• 3D structure

2⃣ The flash drive analogy (why this makes sense)

Think of STM like a USB flash drive:

• You can store many small files

• But only a few large, detailed files

So:

• Simple objects → take up less space

• Complex objects → take up more space

➡ The more detail an object has, the fewer objects you can store.

3⃣ Alvarez & Cavanagh (2004): the key experiment

They used the change detection task (same logic as Luck & Vogel) but varied object complexity.

What participants did

1. Saw a display of objects

2. Short delay

3. Saw a second display

4. Judged same or different

The critical manipulation: object complexity

They compared:

• Simple colored squares

• More complex objects (e.g., shaded cubes with depth and multiple features)

4⃣ The results (this is the key finding)

Memory capacity depended on complexity:

• Colored squares → ~4.4 items

• Shaded cubes → ~1.6 items

👉 Same task, same procedure — just more visual detail.

5⃣ What this means theoretically

Alvarez & Cavanagh concluded:

So STM is limited by:

• Either number of objects

• Or total information load

• Or (most likely) both

6⃣ Why researchers still disagree

There are two competing perspectives:

🟦 Item-based view

• STM stores a fixed number of “slots” (≈4)

• Each slot holds one object

• Supported by Luck & Vogel

🟦 Information-based view

• STM has a flexible resource

• More detail = more resources used

• Supported by Alvarez & Cavanagh

The field has not fully settled this debate.

7⃣ What everyone agrees on

Despite disagreement about how to measure capacity, researchers agree on this:

Whether measured by:

• Items

• Information

• Or both

➡ STM cannot store everything.

8⃣ Why this leads to the next topic (important transition)

Up to now, STM has been treated like:

But this metaphor breaks down because:

• We don’t just store information in STM

• We use it to:

  • Think

  • Reason

  • Calculate

  • Understand language

  • Learn

➡ This is why psychologists moved beyond STM to the concept of working memory.

🔑 One-sentence exam-ready summary

Working Memory: Manipulating Information

1⃣ Why psychologists moved from short-term memory to working memory

The problem with the old STM idea

In the modal model, short-term memory was treated as:

• A temporary storage box

• Where information just sits briefly

But this view couldn’t explain how we:

• Understand sentences

• Follow conversations

• Do mental math

• Read while remembering numbers

• Perform two tasks at once

👉 These activities require storage + active processing at the same time.

2⃣ What is working memory? (core definition)

Baddeley & Hitch (1974) defined working memory as:

🔑 The key word is manipulation.

3⃣ How working memory differs from short-term memory

So:

• STM = holding

• WM = holding + working on

4⃣ Example 1: Understanding language (pizza conversation)

Rachel hears:

To understand this, she must:

• Hold the first sentence in memory

• Process the second sentence

• Link “spinach” to “mushrooms”, not broccoli

This is not just storage.
It’s active integration, which is what working memory does.

5⃣ Example 2: Mental math (43 × 6)

When you solve this in your head, you:

• Store intermediate results (like the 8)

• Carry numbers

• Perform operations

• Update values

If memory were only storage, this would be impossible.

👉 Mental math requires simultaneous storage and processing.

6⃣ The key discovery: dual-task performance

Baddeley noticed something important:

• People can read while remembering numbers

• Or listen to speech while visualizing a route

According to the old STM model:
❌ This shouldn’t be possible (STM would be overloaded)

But it is possible — meaning STM must have separable components.

7⃣ Baddeley’s Working Memory Model (the solution)

Baddeley proposed that working memory has multiple parts, not one box.

The three main components:

🟦 1. Phonological Loop (verbal & auditory information)

Handles:

• Spoken words

• Written words (converted to sound)

• Numbers

• Names

• Language comprehension

Two sub-parts:

• Phonological store: holds sounds briefly

• Articulatory rehearsal process: keeps sounds active by repetition

📌 Example:

• Repeating a phone number

• Understanding your professor’s lecture

🟦 2. Visuospatial Sketchpad (visual & spatial information)

Handles:

• Images

• Shapes

• Locations

• Spatial layouts

• Mental maps

📌 Example:

• Visualizing a route on campus

• Solving a puzzle

• Imagining an object rotating

🟦 3. Central Executive (the “traffic cop”)

This is the control system.

It:

• Directs attention

• Coordinates the two subsystems

• Switches between tasks

• Pulls info from long-term memory

• Suppresses irrelevant information

📌 It does not store information itself.

8⃣ Driving example (why this makes sense)

When you’re driving in a new city:

• Phonological loop → listening to directions

• Visuospatial sketchpad → imagining streets and turns

• Central executive → coordinating, focusing, ignoring the radio

This shows how different systems work in parallel.

9⃣ Why “working memory” is the better term

The key takeaway is:

That’s why modern psychology prefers the term working memory, even though people sometimes still say STM.

🔑 One-sentence exam-ready summary

What the Phonological Loop Does

The phonological loop is the part of working memory that:

• Handles verbal and auditory information

• Represents information as sound, even if it enters visually

It has:

1. Phonological store → holds sounds briefly

2. Articulatory rehearsal process → refreshes sounds by repetition

1⃣ Phonological Similarity Effect

What it is

Classic experiment (Conrad, 1964)

Participants:

• Saw letters flashed on a screen

• Wrote them down in order

Errors were sound-based, not visual:

• “F” mistaken for “S” or “X” (sound alike)

• Rarely mistaken for “E” (looks alike)

Why this matters

Even though the input was visual, the errors were auditory.

👉 This shows that:

• Visual information is recoded into sound

• Verbal memory relies on the phonological store

Real-life example

Phone numbers:

• You see them

• You remember them by repeating their sound

🔑 Takeaway

The phonological similarity effect supports the idea that:

2⃣ Word Length Effect

What it is

Example

Short words (easy):

Long words (harder):

Even though both lists have 8 words, the long-word list is harder.

Why does this happen?

• Long words take more time to pronounce

• Rehearsal is time-limited (about 1.5–2 seconds)

• Longer words “crowd out” the phonological store

Baddeley found:

🔑 Takeaway

The word length effect shows:

This strongly supports the phonological loop model.

3⃣ Articulatory Suppression

What it is

What’s happening?

Repeating “the”:

• Occupies the articulatory rehearsal process

• Prevents rehearsal of target words

• Causes faster decay in the phonological store

Demonstration result

• Normal reading → better recall

• Reading while saying “the, the, the…” → worse recall

Crucial finding

Baddeley et al. (1984) showed:

• Articulatory suppression eliminates the word length effect

Why?

• If rehearsal is blocked, short words lose their advantage

• Both short and long words decay equally

🔑 Takeaway

Articulatory suppression shows:

4⃣ How all three effects fit together

Together, they provide strong evidence for:

• A specialized phonological loop

• Separate from visual memory systems

• Actively maintained through rehearsal

🧪 Exam-style summary sentence

What is the visuospatial sketch pad?

The visuospatial sketch pad is a part of working memory that:

• Holds visual information (shapes, colors, patterns)

• Holds spatial information (locations, orientations, movement)

• Lets you mentally manipulate images (rotate, scan, visualize)

Think of it as your mental whiteboard or mind’s eye.

1. Shepard & Metzler (1971): Mental rotation

What were they testing?

They wanted to know:
👉 Do people actually rotate images in their mind, or do they use some shortcut?

The task

You see two 3D shapes and must decide:

• Same object, just rotated?

• Different objects?

Some pairs are slightly rotated; others are rotated a lot.

Key result

• The more rotated the object is, the longer it takes to decide.

• Reaction time increases linearly with degree of rotation.

What does this show?

People are mentally rotating the object in real time — like turning it in their head.

🧠 Conclusion:
This proves that the visuospatial sketch pad supports active visual manipulation, not just storage.

2. Della Sala et al. (1999): Remembering visual patterns

The task

• Look at a grid of shaded squares for 3 seconds

• Then recreate the pattern from memory

These patterns:

• Are hard to describe in words

• Force you to rely on visual memory, not language

Results

• People could accurately remember about 9 shaded squares

• Even though STM capacity is often said to be 4 items

Why doesn’t this contradict STM limits?

Because people:

• Chunk squares into subpatterns (rows, shapes, clusters)

• Just like chunking words into sentences

🧠 Conclusion:
The visuospatial sketch pad can store more when information is organized visually, not randomly.

3. Brooks (1968): Interference in the visuospatial sketch pad

This part explains why pointing is harder than speaking.

The task

You visualize a large letter F and move around its corners in your mind.

Two versions:

Task 1: Pointing

• Visualize the F

• Point to “In” or “Out” on a table

Task 2: Speaking

• Visualize the F

• Say “In” or “Out” out loud

Which is harder?

👉 Pointing is harder

Why?

Because:

• Visualizing the F = visuospatial sketch pad

• Pointing = also visuospatial

• Two visuospatial tasks compete for the same resource

Speaking uses the phonological loop, so it doesn’t interfere.

🧠 Conclusion:
The visuospatial sketch pad is separate from verbal memory and is disrupted only by visual/spatial interference.

Pulling it all together (exam-ready summary)

The visuospatial sketch pad:

• Stores visual + spatial information

• Supports visual imagery and mental manipulation

• Is involved in tasks like:

  • Mental rotation

  • Remembering patterns

  • Navigating space

• Can be overloaded by other visuospatial tasks

• Is separate from the phonological loop

One-sentence takeaway (perfect for exams)

Big idea first (one sentence)

The central executive is the attention-control system of working memory.
It doesn’t store information — it decides what gets attention, when, and for how long.

What the central executive is NOT

This is important.

• ❌ It does not store words (that’s the phonological loop)

• ❌ It does not store images or locations (that’s the visuospatial sketch pad)

Instead, it manages those systems.

What the central executive DOES

Baddeley describes it as an attention controller. It:

1. Focuses attention

   • Decides what you should pay attention to

2. Divides attention

   • Lets you do two things at once (if possible)

3. Switches attention

   • Helps you stop one task and start another

Think of it as the manager or traffic cop of working memory 🚦

How it fits with Chapter 4 (attention)

In Chapter 4, you learned about executive attention:

• Used for conflict resolution

• Needed for task switching

• Limited and easily overloaded

👉 The central executive is the working-memory version of executive attention

They are essentially describing the same control system, just in different contexts.

Example: Driving while on the phone 🚗📱

Here’s how the textbook example breaks down:

Phonological loop

• Listening to the person on the phone

• Formulating responses

• Understanding language

Visuospatial sketch pad

• Monitoring the road

• Visualizing intersections

• Tracking other cars

Central executive

• Decides how much attention goes to driving vs. talking

• Shifts attention when danger appears

• Tries (often unsuccessfully) to manage both tasks

This is why multitasking breaks down under high demand — the central executive is limited.

Evidence from brain damage: Perseveration

What is perseveration?

Perseveration = getting stuck on one rule, thought, or action.

Example:

• Rule 1: “Pick the red object” → person succeeds

• Rule switches: “Pick the blue object”

• Person keeps picking red, even when told they’re wrong

What does this show?

• The person can perceive

• The person can respond

• But they can’t switch rules

🧠 This means the problem is not memory storage — it’s control of attention

Why frontal lobe damage matters

The frontal lobes are heavily involved in:

• Attention control

• Task switching

• Inhibition of old responses

Damage here → impaired central executive function

This supports Baddeley’s idea that:

Super-clear summary table

Exam-ready definition (use this 🔥)

Working Memory and The Brain

1. Why the prefrontal cortex (PFC) matters for working memory

The PFC is involved in:

• holding information temporarily

• planning and decision-making

• controlling attention

• inhibiting inappropriate responses

These functions match exactly what the central executive does in Baddeley’s working memory model.

👉 So:
PFC ≈ neural basis of the central executive

2. Phineas Gage: what his case shows (conceptually)

What happened

• An iron tamping rod destroyed part of Gage’s frontal lobe

• He survived, but his behavior changed

After the injury

Reports (with some historical debate) described:

• poor impulse control

• difficulty planning

• inappropriate social behavior

• trouble regulating actions

Why this matters for working memory

Even though Gage could still:

• speak

• remember facts

• perceive normally

He struggled with:

• organizing behavior

• using information appropriately

• controlling attention and impulses

➡ This supports the idea that the frontal lobes don’t store information, but control how information is used, which is exactly the role of the central executive.

3. Animal evidence: delayed-response task (this is very important)

The task (what the monkey does)

1. Monkey sees food in one of two wells

2. Wells are covered

3. A delay occurs

4. Monkey must remember where the food was

This task requires:

• holding information during a delay

• no sensory cues available
  → pure working memory

What happens with intact PFC

• Monkeys perform well

• They remember the correct location

What happens when the PFC is removed

• Performance drops to chance (≈50%)

• Monkey guesses randomly

Key conclusion

This directly links the PFC to temporary storage + control, not long-term memory.

4. Developmental evidence: infants and the PFC

Young infants show:

• “Out of sight, out of mind” behavior

• If an object is hidden, they act as if it no longer exists

Why?

• The frontal and prefrontal cortex are not fully developed

• Until about 8 months of age

As the PFC matures:

• working memory improves

• object permanence emerges

➡ Again, this supports the idea that PFC enables maintaining information when it’s no longer visible

5. How this fits into working memory theory

Modern research shows:

• Working memory is not located in one place

• But the PFC coordinates activity across the brain

Exam-ready summary paragraph

Working memory always involves:

1. A stimulus appears

2. The stimulus disappears

3. Information is still needed during a delay

So scientists looked for neurons that could stay active during that delay.

The experiment (Funahashi et al., 1989) — step by step

Task the monkey performed

This is a delayed-response task, designed to isolate working memory.

Step 1: Fixation

• Monkey looks at a fixation point (X)

• This keeps eye position controlled

Step 2: Cue appears

• A square flashes briefly at one location (e.g., upper-left corner)

• The neuron responds a little when the square appears

Step 3: Delay period (THIS IS THE KEY)

• The square disappears

• Nothing is on the screen

• The neuron keeps firing

➡ This firing is not caused by seeing anything
➡ It represents memory of the square’s location

Step 4: Response

• Fixation point disappears

• Monkey moves its eyes to where the square was

If the monkey does this correctly → working memory succeeded.

The key discovery (this is the takeaway)

Certain prefrontal neurons:

• Respond only to specific locations

• Continue firing during the delay

• Stop firing once the response is made

Example:

• Neuron A → fires only for upper-left

• Neuron B → fires only for upper-right

• Neuron C → fires only for bottom-left

Each neuron is tuned to a specific spatial location.

Why this is such strong evidence for working memory

This shows that:

• Memory is actively maintained

• Information does not just “sit” passively

• The PFC uses sustained neural firing to hold information

👉 Working memory = active neural activity, not stored traces

How this connects to theory

This supports the idea that:

• The prefrontal cortex maintains working memory representations

• The central executive has a biological basis

• Working memory depends on ongoing neural firing, not long-term storage

Important contrast (exam gold)

One-sentence exam answer

BIG IDEA 1: Working memory does NOT rely on just one neural mechanism

Originally, researchers thought:

This came from studies like Funahashi et al., where PFC neurons fired throughout the delay.

But newer research shows:

BIG IDEA 2: Two ways information can be held in working memory

🧠 A. Sustained neural firing (classic view)

• Neurons keep firing during the delay

• Information is actively maintained

• Vulnerable to distraction and fatigue

This explains:

• Why attention is required

• Why working memory collapses when distracted

🧠 B. Activity-silent working memory (Stokes, 2015)

This is the newer idea that’s confusing a lot of students — so here’s the intuition:

Think of it like this:

• You don’t need to keep shouting a thought to remember it

• The brain can “store” it by temporarily changing connections

Stokes’s model (in simple steps)

1⃣ Activity state

• Stimulus appears

• Neurons briefly fire

2⃣ Synaptic state (THIS IS THE KEY)

• Firing stops

• Synaptic connections are temporarily strengthened

• No ongoing firing

• Information is stored silently

3⃣ Reactivation

• When memory is needed

• Network becomes active again

• Memory reappears in firing pattern

➡ This is called activity-silent working memory

Why this idea matters

• Explains how working memory can:

  • persist without constant firing

  • be more energy-efficient

• Suggests working memory is dynamic, not static

⚠ Important:
These models are still theoretical, but strongly supported by recent evidence.

BIG IDEA 3: Working memory is distributed across the brain

Working memory is NOT located only in the PFC.

Instead:

• PFC → control & attention

• Visual cortex → visual information

• Auditory/language areas → verbal information

• Parietal cortex → spatial processing

This is called distributed representation.

BIG IDEA 4: Individual differences in working memory = attention control

Vogel et al. (2005): WHY some people have better working memory

Participants:

• High-capacity WM group

• Low-capacity WM group

Task:

• Remember red rectangles

• Ignore blue distractors

Key finding:

• Both groups encoded relevant items equally

• Low-capacity group encoded distractors

• High-capacity group filtered them out

What this means

Working memory capacity is NOT mainly about:
❌ storing more items

It IS about:
✅ controlling attention
✅ blocking irrelevant information

This is a function of the central executive.

The tuning-out conclusion (VERY IMPORTANT)

Research shows:

This links working memory directly to:

• cognitive control

• executive attention

• resisting temptation

• self-regulation

ERP evidence (why it’s convincing)

The event-related potential (ERP):

• Measures how much working memory capacity is being used

• Larger ERP = more information stored

Results showed:

• Distractors increased ERP much more in low-capacity individuals

• Meaning: distractors were taking up working memory space

How everything connects (big synthesis)

One-paragraph EXAM-READY summary