Transcript If-Then Rules

Memory


All learning requires memory
Three stages of memory phenomena



Acquisition
Retention
Retrieval
Taxonomy of Human Memory
Procedural
automatic,
incremental,
unconscious
Motor Skills
Classical
Conditioning
Operant
Conditioning
Declarative
effortful,
conscious
Working and
Reference Memory
Episodic and
Semantic Memory
Reference versus Working Memory
• Reference memory
– long term retention of events, relationships, and
procedures
– associations, rules, skills.
• Working memory
- short term retention, typically relevant only to the
current trial, includes information retrieved
Working Memory in Animals
food
Hunter (1913)
Working Memory or Body Orientation?
food
Delayed Matching to Sample (DMTS)
Comparison
Sample
Comparison
Delayed Matching to Sample (DMTS)
PECK
PECK
FOOD
NO
FOOD
Symbolic Matching to Sample
Symbolic Matching to Sample
PECK
PECK
NO
FOOD
FOOD
NO
FOOD
FOOD
What is Learned in DMTS?
a) General Matching Rule
Pigeon = No! (with few samples)
Cumming & Berryman (1965)
- Trained on Red, Green, Blue
- Failed to transfer to Yellow
b) Specific “If-Then Rules”
Symbolic Matching-To-Sample
- Learned as rapidly as Standard DMTS
Memory Coding
a) Retrospective = Backward Looking
b) Prospective = Forward Looking
Retrospective Code: IF
, Remember
Prospective Code: IF
, Remember
Roitblat, 1980
Confusion Errors?
1. between samples
2. between comparisons
Confusions:
Comparisons > Samples
Therefore:
Prospective Coding
Serial List Learning
Present list of items to subject one at a time
A  B  C  D  E  F
Recall in any order
Serial List Learning
Ask subject to recall or recognize a single item
Recency effect
Accuracy
Primacy effect
A
B
C
D
E
F
Humans:
Testing after a delay produces
a primacy effect
Accuracy
Accuracy
Testing immediately after list
produces a recency effect
A
B
C
D
E
F
What about in other animals?
A
B
C
D
E
F
Radial Arm Maze
How Solved?




Random Choice
Odour Trail
Patterned Responding
Memory*
12-Arm Radial Maze
Choices Prior to First Error
Mean Choices
8
6
4
2
0
1
3
5
7
Trial
9
11
13
12-Arm Radial Maze
Total Number of Errors
10
Mean Errors
8
6
4
2
0
1
3
5
7
Trial
9
11
13
Can rats switch from retrospective to
prospective memory?
12
1
11
2
10
3
9
8
4
7
5
6
Cook et al. (1985)


Rats removed after making 2, 4, 6, 8, or 10
choices
Shifting from retrospective to prospective
midway produces the lowest memory load
(inverted U-shaped error curve)
Cook et al. (1985)
Remember
Places Not Visited
Remember
Places Visited
Memory Coding
a) Active = rehearse relevant information
b) Passive = gradual fading of a memory
trace
Human Forgetting Curve With No
Rehearsal
Pigeon Forgetting Curve
Roberts, 1972
Directed Forgetting
Sample
Remember
cue
Comparison
Forget
cue
peck
don’t
peck
ITI
Forget
cue
Delay
Least
peck
More
peck
Most
peck
Human Reference Memory




Duration (relatively long-term)
Capacity (relatively large)
Forgetting (details lost, gist remembered)
Requires Consolidation
Retention of Fear Conditioning
Clark’s Nutcracker





Food Storing Bird
About 5,000 Caches
20 x 20 KM Area
9-month
Buried Under Snow
Sarah Shettleworth
Results



Birds recovered previously cached seeds and
made few errors
Didn’t find seeds hidden by experimenter
Didn’t return to the same site if first storing
episode is followed by a second storing
episode
Summary of Animal Memory

Prospective and Retrospective 
Active and Passive

Reference memory

Duration and Capacity

Forgetting and Consolidation 
Working Memory
Do Animals have Episodic Memory?

Episodic Memory


Conscious Recollection
Dated Personal Memory (what, when, and where)
Western Scrub-Jay (Nicola Clayton)
Clayton’s Results
Metamemory in Rats?

Knowledge of the state of one’s own memory


for example, memory strength
Foote and Crystal (2007)




Duration of noise sample, 2.00 to 3.62 = Left
Duration of noise sample, 4.42 to 8.00 =Right
Choice to continue → memory test, large reward
Choice to bail-out → no test, small reward
Foote and Crystal (2007) Procedure
Foote and Crystal (2007) Results
Problems


Only 2 of 3 rats showed positive results (5
others always bailed or always decided)
Maybe they learned to bail with feedback on
the “close” duration values?