Building Memories: Neural Activity During Learning Predicts Later Remembering

1
Neuroimaging of Memory Systems
Anthony D. Wagner Department of Psychology
Neurosciences Program
2
Multiple Memory Systems
LTM
Corkin et al (1997)
Declarative
Nondeclarative
Episodic Semantic Skill Learning
Priming Conditioning (events)
(facts)
Squire (1995)
3
Agenda

• fMRI measures of episodic encoding
• fMRI correlates of episodic retrieval
• Hippocampal function
• MTL Cortical function
• Issues of necessity

4
Episodic Encoding
Episodic Memory Conscious memory of life events
Strategy 1 Vary some factor that is known to
modulate later memory performance and
characterize how the neural response
differs. Strategy 2 Hold task constant and back
sort the encoding events into those subsequently
remembered and those subsequently forgotten.
5
Strategy 1 Task Manipulations at Encoding

• Binary comparisons
• full- vs. divided-attention during episodic
  encoding (e.g., Shallice et al., 1994)
• deep vs. shallow processing during episodic
  encoding (e.g., Kapur et al., 1994)

6
Strategy 1 Task Manipulations at Encoding
Left PFC and MTL activation associated with
deeper encoding
(Wagner et al., 1998)
7
Issues to Consider

• 1) Encoding success or Encoding content?

Kirchhoff et al. (2000) J Neuroscience
8
Building Memories Multiple Circuits
Event content and attentional orientation
influence which PFC-posterior neocortical circuit
contributes to encoding
9
Issues to Consider

• 1) Encoding success or Encoding content?
• 2) Subtractive logic assumption of pure
  insertion
• conditions may differ along multiple dimensions
• requires correct task analyses
• assumes processes do not interact

10
Measuring Memory FormationThe Subsequent Memory
Paradigm
Approach Use fMRI to identify brain events
during an experience that predict whether the
experience will be later remembered or forgotten.
Sanquist et al., 1980 Neville et al., 1986 for
review, see Rugg, 1995 Wagner et al., 1999
11
Measuring Memory FormationThe Subsequent Memory
Paradigm
Paller Wagner (2002) Trends Cog Science
12
Measuring Memory FormationThe Subsequent Memory
Paradigm
Paller Wagner (2002) Trends Cog Science
13
Neural Correlates of Verbal Encoding Subsequent
Memory Effect
Left Inferior PFC
14
0
2
4
6
8
10
12
Left Parahippocampal Cortex
4
3
2
1
0
-1
0
2
4
6
8
10
12
14
Peri-Stimulus Time (sec)
Wagner et al. (1998) Science
14
Continuous Encoding Effects Less Subject to
Concerns about Pure Insertion
Ranganath et al. (2004)
15
Issues to Consider

• 1) Intentional vs. Incidental encoding
• Psychology Intent is not necessary
• Imaging Intent can result in temporally
  displaced
• cognition
• 2) How to understand subsequent memory effects
• Item effects Need to rule out
• Why were some items better remembered?
• 3) Effects interact with the nature of the
  retrieval test

16
Multiple Subsequent Memory EffectsHippocampal
vs. MTL Cortical Learning
Do distinct encoding computations support
subsequent conjunctive memory vs. item memory?
LOGIC Identify encoding predictors of
subsequent context recollection vs. item
recognition
17
Sort Encoding Data by Subsequent Item Memory and
Context Recollection
Davachi, Mitchell, Wagner (2003) PNAS
18
Dissociable MTL Encoding Mechanisms
ItemConjunct
19
Agenda

• fMRI measures of episodic encoding
• fMRI correlates of episodic retrieval
• Hippocampal function
• MTL cortical function

20
Medial Temporal Lobe A Convergence Zone
Squire Kandel (1998)
21
Functional Dissociations within MTL
22
Memory Retrieval Hippocampus and Remembering
Remember
Know
Correct Reject.
Miss
Eldridge et al. (2000) Nature Neurosci
Yonelinas et al., (2005) J Neurosci
23
Issues to Consider

• Encoding contaminants during retrieval
• Psychology Retrieval is a powerful encoding
  event
• MTL is sensitive to stimulus novelty at
  encoding

0.5
0.4
0.3
Signal Change
0.2
0.1
0
-0.1
0
2
4
6
8
10
12
14
Peri-stimulus time (s)
Kirchhoff et al. (2000) J Neuroscience
24
Issues to Consider

• Encoding of foils may mask retrieval responses to
  targets

CRs and Misses are more novel than Know
items Is it possible that MTL novelty encoding
responses to these items masks MTL retrieval
responses to Know items?
Remember
Know
Correct Reject.
Miss
25
Issues to Consider

• Foils on retrieval tests are encoded during the
  retrieval task

Stark et al. (2003)
26
Memory Retrieval Hippocampus and Recollection
Context Recollection gt Item Recency
Dobbins et al. (2003) Neuropsychologia
27
Functional Dissociations within MTL
28
MTL Cortical Responses at Retrieval Repetition
Suppression

• MTL Cortex (Perirhinal Cortex)
• novel gt repeated
• 25 of neurons
• single trial
• early onset (100-150 ms)

e.g., Brown Aggleton, 2001, Nat Rev Neurosci
Eichenbaum, 2000, Nat Rev Neurosci
29
MTL Cortex (fMRI) Familiarity-Related Response
Decrease
Baseline gt Retrieval
Gonsalves et al., 2005, Neuron
30
MTL Cortex (fMRI) Familiarity-Related Response
Decrease
31
Issues to Consider

• 1) Baseline task
• Unconstrained baselines (e.g., fixation or
  rest) may allow subjects to engage in cognition
  that drives the MTL
• Can lead to null results
• Baseline is arbitrary look for regions that
  are sensitive to memory status even if levels are
  below baseline

(Stark Squire, 2001)
32
Issues to Consider
Relative to Easy Relative to
Difficult
Rest gt Visual pattern detection (X or T in
noise?) (Stark Squire, 2001)
Difficult Easy
(Law et al., 2005)
33
Issues to Consider

• 1) Baseline task
• Unconstrained baselines (e.g., fixation or
  rest) may allow subjects to engage in cognition
  that drives the MTL
• Can lead to null results
• Baseline is arbitrary look for regions that
  are sensitive to memory status even if levels are
  below baseline
• 2) Spatial resolution
• Subregions of hippocampus may functionally
  differ

34
High-Resolution fMRI of Human MTLFunctional
Dissociations
Faces
CA2/3/DG
Collateral Sulcus
Subiculum
CA1
PRc
Scenes
Preston et al. (in prep)
35
CA Fields vs. Subiculum
Zeineh et al. (2003)
36
Agenda

• fMRI measures of episodic encoding
• fMRI correlates of episodic retrieval
• Hippocampal function
• MTL Cortical function
• Issues of necessity

37
Multiple Memory Systems
LTM
Corkin et al (1997)
Declarative
Nondeclarative
Episodic Semantic Skill Learning
Priming Conditioning (events)
(facts)
Squire (1995)
38
Delay Conditioning and the MTL
Tone or Static Noise
Air Puff follows Tone
Con - Aware
Con - Unaware
Amnesics
(Clark Squire, 1998)
Block of 20 Trials
39
What neuroimaging does NOT tell us

• Correlational data -- does not address necessity
• e.g., delay conditioning and MTL (Blaxton et
  al., 1996)

• Fig. 4. Changes in activation across conditioning
  trials for regions in left frontal cortex, left
  hippocampal formation, and cerebellum.
  Activation in cerebellum and left hippocampal
  formation increased as learning trials
  progressed, but the left frontal cortex showed
  the opposite trend, actually becoming less
  activated as the CS-US association was learned.

40
Neuroimaging of Memory Systems