Working Memory

Working Memory

Active short-term memory – Maintenance of task-relevant informationonline. Keeps relevant information available.

Like RAM in a computer

Close correspondence between working memory and consciousness

Delayed response tasks are used to WM in animals

Spatial delayed response

Object delayed response

also alternation tasks

In all cases, a response is based on some previous event that is relevant on the current trial, but not necessarily the next.

Sample

What Delay

Test Array

Where Delay

What-Then-Where Task

4 objects and 4 locations were used

Choice

Fixation (2000 msec)

Sample (500 msec)

Delay (1000 msec)

Test (500 msec)

S1

S2

S3

C1

C2

C3

Delayed paired associate task

Important features of WM tasks

A cue is followed by a delay and then two or morechoices

Delay is relatively short (about a second to about a minute).

Small number of cues that are used repeatedly. A singlecue is only relevant for one trial at a time and animalsmust remember

Delay tasks require that:

1. Animals discriminate among cues

2. Maintain memory across a delay

3. Evaluate choice phase and make correctdecision

Brain areas involved in WM?

Today’s take-home message:

Working memory is not simply short-term memory.It is an active process in which subjects must select andmaintain information. It is under volitional control.

Two fundamental observations of behavior duringa typical WM task (delayed match-to-sample)

1. Performance increases with increases in sample

Correct choice

Sample Delay Test

Time

2. Performance decreases with increases in the delay

(Much of this work has been conducted in pigeons)

Also:Match rule retained inreference memory.

The copy/trace model was meant to highlight what WMis NOT.

Other observations of DMS performance:

1. Changing the illumination of the testing box during retention intervalinterferes w/ performance. (doesn't matter whether light->dark dark->light)  This depends on degree of "surprise". It will habituate. Expected events have no effect on performance.

2. Surprising samples are more memorable.

Suggests that animals actively process info about the SM during delay.When surprised, attentional resources drawn to event, less left for actively maintaining the sample.

Another example: Directed Forgetting

Next: Neurophysiologicalevidence for WM as anactive process

Sample

What Delay

Test Array

Where Delay

What-Then-Where Task

4 objects and 4 locations were used

Choice

'What" Delay Activity in PF Cortex: Single Cell

Time from sample onset (msec)0 500 1000 1500 2000 2500

Spikes per second

20

30

40

50

60

70

80

90

100 'What' Delay 'Where' Delay

Sample “A”

Sample “B”

a40_1_01

'What' and 'Where' Delay Activity: Single Cell

'What' Delay

Time from sample onset (msec)

0 200 400 600 800 1000 1200 1400

Spi

kes

per s

econ

d

20

40

60

80

100

Sample A

Sample B

a43_1_32

'Where' Delay

1400 1600 1800 2000 2200 2400 2600 2800

Down

Up

A monkey test of attention

Sample

Array Trials

Delay Test stimuli

Nonmatch Match

Remember the identity and location of only the relevant object

Cue Trials

750 msec 1500 msec Bar release

Neural correlate of awareness in the prefrontal cortex

Time from sample onset (msec)

0 1000 2000

Spikes per second

0

50

100

Time from sample onset

0 1000 2000

Sample Test

“A” alone “B” alone

Neural correlate of awareness in the prefrontal cortex

Time from sample onset (msec)

0 1000 2000

Spikes per second

0

50

100

Time from sample onset

0 1000 2000

Sample Test

Attend “A” “A” alone “B” alone

Neural correlate of awareness in the prefrontal cortex

Time from sample onset (msec)

0 1000 2000

Spikes per second

0

50

100

Time from sample onset

0 1000 2000

Sample Test

Attend “A” “A” alone Attend “B” “B” alone

WM is not a passive copy of a sensory stimulusinstead an active process  So, in contrast to the copy/trace model, WM is:1. Active - Ss actively select, encode and maintain, what is heldin WM.2. Not a simple match rule. Simple maintenance of record of samplethen application of a matching rule. Rather something more complexis going on. One thing that may be happening is that instead ofa matching rule, Ss develop an expectation, they anticipate whatis going to happen and base their behavior on this expectation

What is info is maintained in the delayof WM tasks?

Retrospective versus Prospective processing

Correct choice

Sample Delay Test

Time

There is a considerable amount of evidence that animals tend to relyon a prospective code for performance in delay tasks.

In symbolic DMS, increasing the delay between sample and choice stimuli causes an increase confusions between similar choice stimuli, not similar samples.

Fixation (2000 msec)

Sample (500 msec)

Delay (1000 msec)

Test (500 msec)

S1

S2

S3

C1

C2

C3

Fixation (2000 msec)

Sample (500 msec)

Delay (1000 msec)

Test (500 msec)

S1

S2

S3

C1

C2

C3

Delayed paired associate task

Reaction Time [msec]250 300 350 400 450

Freq

uenc

y of

occ

uren

ce [%

]

0

5

10

15

20

S1S2

S3

C1

C2C3

S1 à C1C1 à C1

S2 à C2C2 à C2

S3 à C3C3 à C3

Time [msec]0 500 1000

Pro

spec

tive

Inde

x

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08

0.10

***

* P<0.01

Conclusions

Working memory is an active process. It is under volitionalcontrol and can be contrasted with passive or automaticforms of memory.

Working memory is prospective. It is geared toward the future, notmerely a record of past events.