Transcript Mathematical Models

Memory and its disorders Models
of memory function
Chris Moulin
School of Psychology
University of Leeds
[email protected]
Administration
• Slides are here:
Plan
Types of Models
Cognitive, box and arrow models
Mathematical models
Computational models
Neural models
Neuroimaging models
•
•
•
•
Working Memory
Procedural/Implicit/Non-declarative memory
Autobiographical Memory
Episodic and Semantic Memory
Types of Model
Mathematical models
Computational models
Neural models
Neuroimaging models
Cognitive, box and arrow models
Mathematical Models
• Signal detection theory
I spy in a signal detection paper:
ROC curves
d’ (D-Prime)
Boring theories
Hits, misses, correct rejections and false positives
• Powerful Model for recognition memory
Theoretical Assumptions
• Trace strength model based on normal
distribution
• There is signal and there is noise
Recollection in Rats
Computational Models
• E.g. Connectionism
• A way of modelling the brain?
• You can train these networks to do things
Neural Models
• E.g. Long Term Potentiation
LTP
• Brain made up of billions of neurons, which are
connected by synapses.
more LTP
• Repeated activity of the neuron causes
synapses to become potentiated - meaning
they fire easier.
• Networks of neurons can be activated when
only one neuron in that group is stimulated.
even more LTP
• Further research found that after continued activity of
neurons new synapses are grown to connect to other
neurons.
• So networks of neurons can become massive e.g.
one neuron can be connected to as many as 1000
other neurons.
• Associations in LTP are similar to associations in
classical conditioning.
Neuroimaging Models
• E.g HERA.
Converging Evidence
Cognitive theory says
that there is ability A
and ability B and they
are separate
A
B
If this is true, then they
should be in separate
areas (but it is not
disproved if they’re not)
Neuroscience Tests
cognitive models
HERA
• Hemispheric Encoding/Retrieval Asymmetry
Encoding – taking information in
Retrieval – getting that information back
• For verbal materials:
Left frontal lobe – encoding
Right frontal lobe – retrieval
Encoding
controls
Retrieval
controls
HERA – so what?
• HERA not generally supported by patient
studies.
• Left sided activation is prevalent for any verbal
material that is later remembered
• Largely PET-based and doesn’t extend to fMRI
• Does it help us understand memory and
memory problems?
E.g. Aging
Box and Arrow models
•The majority of memory models are this
•But converging models are the most powerful
Memory Refresher
• Memory is not a monolithic entity.
MEMORY
Declarative
/Explicit
Long Term
Episodic
(Autobiographical
/ past experience)
Short Term
and
Working
Memory
Semantic
(knowledge /
facts)
Procedural
/ Implicit
Long-term memory systems
Declarative memory
(Explicit)
Episodic
memory
Semantic
memory
Non Declarative memory
(Implicit)
Procedural
memory
Procedural/Implicit
• Well learned skills tend not to decay too much
• Cognitive measure: Priming
• Laver & Burke (1993) Meta analysis: if anything,
older adults show more priming.
• Light & Singh (1987). Nice memory test:
TRUMPET
DIAMOND
BANANA
CHERRY
SILVER
MELON
CLOUD
Light & Singh, 1987
•
•
•
•
Mel ____
Sil_____
Tru ______
Clo____
Che_____
Dia ____
Tested either with implicit stem completion or explicit
Only difference is instruction
Old = Young on implicit
Old impaired when instructed to use a word they’d previously
seen
Explicit Memory
The early days
Explicit Memory
Ebbinghaus’s forgetting curve
Reasons for forgetting
• decay of memory trace
• retrieval problems - not lost but misplaced
• interference between memory traces
• Interference is important in Stimulus-Response
• E.g. Hugo Munsterberg
Retroactive & proactive
interference
• Interference - cornerstone of behaviourist
approach to human verbal learning
• RI - new learning interferes with recall of old
learning
e.g. old and new mobile phone functions and buttons
• PI - old learning interferes with current learning
e.g. friend’s new girlfriend’s name
Tests of RI & PI
Group
Exp
Learn
Cat-Tree
Learn
Cat-Dirt
Test
Cat-Dirt
Cont
-
Cat-Dirt
Cat-Dirt
Group
Exp
Learn
Cat-Tree
Learn
Cat-Dirt
Test
Cat-Tree
Proactive
Retroactive
Cont
Cat-Tree
-
Cat Tree
Modern Ideas
• The basics.
Phases of memory
encoding
storage
retrieval
All three of these must be intact
Retrieval
• Explicit / Implicit retrieval
Difference in awareness in retrieval
• Ecphoric / strategic retrieval
(Moscovitch, 1992)
Ecphoric retrieval: effortless retrieval on the base of a simple cue
Strategic retrieval: effortful retrieval. Memory strategies required
Short term/Long term memory
External input
Sensory memory
Large capacity storage
Short time : 2s
Short-term memory
Small capacity
Short time: 30s
Long-term memory
Enormous capacity
Long time
Atkinson and Shiffrin’s model (1969)
Primary/Short-term
• E.g. the recency effect – based on the
phonological loop
• Generally, Primary memory is not impaired by
age
young
Recall (Prop. Correct)
1
0.9
old
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2
3
4
5
6
Serial Position
7
8
9
10
Working Memory
• A new model : Baddeley and Hitch (1974)
'Perhaps short-term memory had no function other than to
keep experimental psychologist amused ? it that were so, we
would rather amuse ourselves in other ways' Baddeley
Working memory
• Think of working memory as short-term
memory with some clever stuff added.
• Simultaneous storage and manipulation
central
executive
phonological
loop
visuospatial
sketchpad
Working Memory
Central executive
Visuospatial
sketchpad
Episodic buffer
Long-Term Memory
Baddeley (1986, 1997)
Phonological loop
Some critical themes
•
•
•
•
•
Verbal Learning Approach - Interference
Memory as process - Levels of processing
Technology affects models
Neuropsychology tests models
Lack of congruity between models
• But models guide neuropsychology
Semantic dementia
Errorless learning