Transcript Encoding

PS4529/30
Applications
of Cognitive
Neuroscience
Consensus View of Long-Term Memory
ENCODING
RETRIEVAL
Seeing Word
MTL
MTL
Hearing Word
MTL
MTL
A Specific Example
• The constructive memory framework (CMF)
• Schacter, DL, Norman, KA, and Koutstaal, W.
(1998). The cognitive neuroscience of
constructive memory. Annual Review of
Psychology, 49, 289-318.
• Invokes multiple brain regions
• Some involved in encoding and retrieval
• Some involved in either encoding or retrieval
• Comprising multiple functions that must interact
dynamically with one another
CMF Neuroanatomy
• The hippocampal formation
 ‘Indexing’ of episodes: exactly how is unknown
 Necessary both for encoding and retrieval
 Damage leads to dense retrograde and anterograde
amnesia
• The frontal lobes
 Strategic control over memory: exactly how is again
unknown!
 Damage leads to confabulations, delusions,
heightened false memory, source amnesia
• The entire ‘association’ neocortex
 Representation of experienced content
 Damage should lead to loss of specific content of
prior episodes
CMF Retrieval Functions
 Retrieval
‘focus’
 Access
to the records of attended
information via a retrieval cue (by
hippocampal pattern completion)
 Inhibition
of irrelevant information
 Re-activation
of episodic content (held
in the neocortex)
 Monitoring/evaluating
retrieval products
(prefrontally mediated)
How Many Experiences Have you Had?
1000000000
Number of episodes (log)
100000000
10000000
1000000
100000
10000
1000
100
10
1
HOUR
DAY
WEEK
MONTH
TIME
YEAR
DECADE
Conway’s SMS Model
The Self Memory System (SMS) has
two principle components:• 1. Autobiographical knowledge base
– organised specifically to support our
sense of self
• 2. The (working) Self
– comprises a goal hierarchy, and various
other internal mechanisms
‘Self’
Autobiographical Knowledge Base
Self-related
semantic knowledge
Abstracted from specific
experiences
Episodic Memory tied to
specific experiences
(e.g. the CMF)
Encoding
Consolidation
Retrieval
Take a peek inside yourself…
Goal
Hierarchy
Self
Episodic Memory
(CMF)
Autobiographical
Knowledge Base
Conceptual
Self
semantic
knowledge
Key point: the SMS system is ‘goal-driven’
The SMS greatly extends the CMF
• Conway: “all daily experiences are
destined to be forgotten”
– Unless they support longer-term goals
• In the short term, accurate memories are
vital
– Where did I leave my keys
• In the long-term, coherence (between
goals) is vital
– The ‘Husband-Hermit’ or ‘Saint-Sinner’ dilemma
SMS Goals
• Short term (e.g. daily)
– Take the car for a service…
– Find the keys…
– Post the letter…
Overriding principle: accuracy!
• Long-term
–
–
–
–
–
Get a job
Learn how to drive
Buy a house
Potential for
Become a solitary religious hermit
Be a loving husband
conflict!
• Overriding principle: coherence
But: there is an accuracy-coherence trade-off!
How is the trade-off achieved?
• The goal
hierarchy
maintains a
stable and
coherent set
of short and
long term
goals
Goal
Hierarchy
Eat and drink (everyday)
Keep warm (everyday)
Have a conversation (most days)
Watch TV (particular times)
Find the car keys (in 5 minutes)
Avoid tripping up (when I walk)
Post the letter (sometime today)
Dentist appointment (this week)
Revise (next month)
Obtain graduation ball tickets (next few
months)
Find a less annoying partner (yesterday!)
Get a 2:1 (next couple of years)
Loose weight (before going on holiday)
The SMS: key points
• New memories are not formed ‘automatically’ from our
experiences
– But, experiences are always encoded (e.g. as per the CMF).
• An ABM is formed (and retrieved) only when the (working) self
interacts with the autobiographical knowledge base
– Such interactions are entirely goal-driven
• Hence, specific experiences will be forgotten unless they
relevant to a goal - within the goal hierarchy
• Stable self-image emerges from the coherence imposed by the
goal hierarchy, perhaps at the expense of accuracy
‘Self’
Autobiographical Knowledge Base
Self-related
semantic knowledge
Abstracted from specific
experiences
Episodic Memory tied to
specific experiences
(e.g. the CMF)
Encoding
Is this: (1) A scientifically acceptable and
(2) A forensically useful model?
Consolidation
Retrieval
Can ERPs reveal exactly what is
happening in the brain while
people remember their past?
Constructive Memory Framework
Stimuli
Time
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
‘selective attention’
Retrieval Perception/
Pattern
completion/
attention
Binding
cue
onset
Ecphory/
inhibition
Monitoring
Focussed Search
Retrieve / Inhibit
Monitor
Stimuli
Time - 0
1
2
3
4
Retrieval failure
5
6
7
8
9
Retrieval success!!
ERP correlates of retrieval from
long-term memory
Stimuli
Time
0
0.1
0.2
0.3
0.4
Familiarity?
0.5
0.6
0.7
F-N400
Implicit Memory?
Ecphory?
LP effect
Monitoring?
Donaldson, Allan and Wilding (2003)
Mecklinger (2000)
Rugg and Wilding (2000)
RF effect
Content only (versus failure)
Content and Context
LP effect magnitude X content relation
Using ERPs to investigate the notion
of encoding-retrieval overlap
1. By manipulating the content of
what is encoded and retrieved.
2. By manipulating the timing of
encoding and retrieval, to make them
coincide.
1. Manipulating Content
• Operationally define different classes of study
episode
• Record EEG when instances of each class of
episode are recollected
• Form ERPs to each class of recollected episode
• Contrast the magnitude and topography of
ERPs for each class of recollected episode
Encoding and Retrieval in vivo…
Encoding
Retrieval
MTL
MTL
Visual
Olfactory (Gottfried et al, 2004)
Auditory
and within ‘sensory domain’ too
(Woodruff et al., 2005)
TIME
Encoding
Retrieval
MTL
MTL
Do ERPs reveal
modality specific retrieval processes?
• Subjects SAW and HEARD words at study
• Performed a word-stem (e.g. MOT__) cued recall task
• ERPs were formed to stems completed with
 Studied SEEN items
 Studied HEARD items
 Unstudied NEW items
• ERP retrieval effects for each sensory modality: SEEN – NEW difference
 HEARD – NEW difference
Allan, Robb and Rugg (2000),
Neuropsychologia, 38 1188-1205.
No!
ERPs are insensitive to
differences in modality at retrieval
Recall visual episode
As retrieval
begins…
As
retrieval
ends…
Recall auditory episode
ERP Modality Experiment: Conclusions
• Multiple retrieval processes, active at different times
– Onset ~ 0.5s after retrieval cue!
• Retrieval of ‘visual’ and ‘auditory’ episodes involves
common processes. No evidence for modality specific
retrieval processes.
• ERPs reflect a ‘core component’ of retrieval?
– Changes in neocortical activity driven by the
Hippocampus during early stages of retrieval (prior to
modality specific activations)?
– Or: attention to retrieval products?
Episodic Memory Mechanisms
Semantic
Records
Perceptual
Records
Semantic
Records
Context
Binding
Encoding
Attentional
Control
Perceptual
Records
Context
Binding
Storage
Consolidation
Mechanisms
Retrieval
Attentional
Control
2. Can we simultaneously
encode and retrieve?
• Gain precise control over the relative timing of events experienced in
different modalities.
• Stress the system by forcing it to handle very rapidly changing
inputs, to reveal what the temporal limits are.
• Examine resulting performance behaviourally
• And use high temporal resolution neurophysiological data to expose
the underlying functional states
Allan and Allen (2005), Journal of
Neuroscience, 25, 8122-9130.
How Many Experiences Have you Had?
1000000000
Number of episodes (log)
100000000
10000000
1000000
100000
10000
1000
100
10
1
HOUR
DAY
WEEK
MONTH
YEAR
DECADE
TIME
Does encoding temporarily stop when retrieval occurs?
3-Phase Dual-task Paradigm
Phase 1
E
E
Phase 2
R
R
E
E
Phase 3
E
Key
Auditory
Visual
R
R
R
+/- 200msec period of stimulus onset asynchrony E == encode (animacy task)
(SOA) jitter, in 10 time bins (41 – 194msec)
R == retrieve (old/new recognition)
Stimulus-Onset Asynchrony (SOA)
Encode time-line begins
Time
controlled ‘jitter’
WILD
Retrieve time-line begins
Expt. 1 SOA range: 50-200msec
Expt. 2 SOA range: 50-2000msec
Dual-task Performance (Expt. 1)
80
70
% correct
60
50
40
30
20
10
0
phase 2
phase 3
Full Hit
71.85
62.66
DA Hit
72.03
49.38
Full CR
79.21
71.04
DA CR
76.88
Effect of ‘ignoring’ retrieval cues
Retrieval under full attention
Retrieval under distracted
attention
LP effect minimal/absent under DA, replaced by F-N400
but no reliable topographic differences (Allan and Allen, 2005)
Conclusions
• Encoding stabilised at a temporal gap of ~600msec
(see Expt. 2), i.e. just as the ERP effect begins.
– retrieval cue processing is complete.
– neocortical trace reactivation has commenced.
– so ‘automatic’ encoding of experience can begin again?
• Retrieval shows a subtle alteration towards reliance on
familiarity
• Mode-shifting between encoding and retrieval in human
memory is relatively sluggish
•The LP effect may reflect the attention paid to retrieval
products, not the representational nature of those
products
Next week Do these ERP effects objectively indicate the
presence / absence of an episodic memory?