Transcript Slide 1

PVLV Model of
Phasic Dopamine Learning
R. O’Reilly
T. Hazy, J. Reynolds, G. Frank
Temporal Difference Dopamine
Reward Model
Brain Dopamine Reward Response = Reward Occurred – Reward Predicted
Schultz, Dayan & Montague 1997
DA ↑
Unexpected Reward
DA ↑
DA
↔
Expected Reward
DA ↑
DA
↓
Unexpected No-Reward
Phasic Dopamine Firing
• Transition from Reward/US to CS onset (Schultz et al)
• TD model says why, but not how (neural mechanisms)
• And TD not a good fit for growing set of data..
PVLV Model
• Central Nucleus of the Amygdala (CNA) = CS -> DA
• Ventral Striatum (NAc, patch) = US -> no DA
4
PVLV Model
• PV = Cancelling US burst
• LV = Driving CS burst
• No chaining: just simpler
delta-rule/Rescorla Wagner
5
PVLV Predictions
• CS – driven DA is dissociable from US – driven:
– Conditioned orienting & autoshaping to a CS =
CNA, but not to US
– CS DA is not subject to blocking effect
– CS DA cannot drive further CS learning: 1st order
(CNA) is dissociable from 2nd order (BLA), and no
3rd order
• Time is just another input; no temporal
chaining as in TD
PVLV Drives PBWM
ROI Activation Analysis after Full Brain
Statistical Analysis
Guido K.W. Frank, M.D.
The Children’s Hospital,
University of Colorado Denver
Receiving Sucrose (US)
Unexpected
AN greater than CW
p=0.005
5vox
uncorr.
SVC p<0.02, FEW, FDR, Cluster
Receiving Sucrose (US)
Unexpected
 Time Activity Curves
VTA
CNA-L
CNA-R
AVS-L
AVS-R
VTA
CNA-L
CNA-R
AVS-L
AVS-R
Anorexia
Nervosa
Control
Women
Influence of Emotion/Reward &
Punishment on EF:
Incentive Stroop Task
Jeff Spielberg, with Wendy Heller, Gregory A. Miller, Laura Crocker, Stacie
Warren, Christina Murdock-Jordan, post-doc Dave Towers, Brad Sutton &
Tracey Wzalek at BIC, and Center colleagues Marie Banich & Randy O’Reilly
Department of Psychology and Beckman Biomedical Imaging Center, University
of Illinois at Urbana-Champaign
Influence of Emotion/Reward &
Punishment on EF
• Multiple ways to conceptualize emotional processes
psychologically
• Circumplex models, valence, bipolarity
– Valence/arousal, positive affect/negative affect
– Valence-based dichotomies prevalent (pos. vs. neg.
emotionality/temperament/extraversion)
• Fundamental motive systems that underlie behavior
– Appetitive/approach/behavioral activation/incentive motivation
– Defensive/withdrawal/behavioral inhibition/aversive motivation
Influence of Emotion/Reward &
Punishment on EF
• No comprehensive model(s) of brain function or structure
integrating roles or mechanisms of brain regions implicated in
emotion:
• Left & Right DLPFC: positive/negative affect
• dACC: anxiety
• Amygdala: fear
• Nucleus accumbens: reward/punishment
• Posterior ACC: emotional autobiographical memory
• Orbital frontal cortex: reward punishment
• Distinct lines of research, most not integrated with each other
Influence of Emotion/Reward &
Punishment on EF
• Lateralization a predominant feature of models of PFC
organization for emotion
• Divided according to superior/inferior lines
– Davidson motivation model, also Harmon-Jones, Coan, &
Allen, others
• Approach/withdrawal mapping onto left vs. right DLPFC
– Heller, Miller, Banich, and others’ valence/arousal model
• Positive/negative valence mapping onto left vs. right DLPFC
• Adds emotional arousal & right parietal activity
Prefrontal Lateralization for Emotional Processes
Approach
Motivation
Withdrawal
Motivation
L
Positive
Emotion
R
Negative
Emotion
Influence of Emotion/Reward &
Punishment on EF
L
R
Distinguishable, lateralized prefrontal areas sensitive to…
Positive valence
L
Approach Motivation
Attachment
Trait Anger
L
R
Anger Out
Anxious Apprehension
Anxious Arousal
Anhedonic Depression
(From Engels et al., 2007; Herrington et al., 2005; Herrington et al.,
2009; Spielberg et al., 2008; Stewart et al., 2008;
all regions depicted replicated in at least two studies)
R
Influence of Emotion/Reward &
Punishment on EF
• Patterns of lateralization
for emotion tend to be
reversed for orbital
frontal cortex (OFC)
• Positive/negative
valence mapping onto
right vs. left OFC
• Note LDLPFC spot!
deAraujo, Rolls, Kringelbach, McGlone, & Phillips, 2003
Influence of Emotion/Reward &
Punishment on EF
• Incentive Stroop Task:
– Engages motivational systems, emotional systems, and executive
functions simultaneously
– TASK
• Press a button ASAP when a word appears
– AFFECTIVE CONTEXT COMPONENT OF THE TASK
• IGNORE the meaning of the word, which can be positive, negative, or
neutral
– Design allows us to examine the effect of the emotional content of the
irrelevant information (affective context) on the ability to ignore that
information
Influence of Emotion/Reward &
Punishment on EF
• MOTIVATIONAL COMPONENT OF THE TASK
– Before each word is shown, participants see a cue which tells them whether
they will win and/or lose money depending on how fast they push the button
– If they push the button fast enough they get the positive outcome on that trial
(i.e., win money or avoid losing money)
– If they don’t push the button fast enough they get the negative outcome on
that trial (i.e., lose money or miss winning money)
– On some trials, they neither win nor lose money
• Allows us to examine the effect of anticipating rewards and punishments
on the ability to ignore irrelevant information
For dollar sign on left:
if it’s green, can win money if push button fast enough
if it’s grey, can’t win money on that trial
For dollar sign on right:
if it’s red, can lose money if don’t push button fast enough
if it’s grey, can’t lose money on that trial
Fast enough = win money
Fast enough = win money
Too slow = lose money
Too slow = do not win money
Fast enough = do not lose money
Do not win or lose money
Too slow = lose money
regardless if they are fast or slow
Incentive Stroop Task
Cue
ISI
Emotional word
ISI
Feedback
Influence of Emotion/Reward &
Punishment on EF
• Recruiting 3 groups of subjects according to PANAS
scores
1) High positive, low negative affect
2) High negative, low positive affect
3) Low negative, low positive affect
• Participants are run in counterbalanced EEG/fMRI
sessions
• SCIDs & assessment of EF components via
neuropsychological tests
Influence of Emotion/Reward &
Punishment on EF
• Preliminary behavioral findings:
– Emotional content of the irrelevant information affects ability
to ignore that information
• Pleasant or unpleasant words elicit slower RTs than do neutral
words
• Emotional words thus harder to ignore
– Motivational context also affects the ability to ignore irrelevant
information
• RT faster on trials in which reward or punishment is possible
• Thus, possible rewards and punishments make it easier to ignore
irrelevant information
– Findings indicate effects of both affective context and
motivation on executive function (as measured by RT)
Influence of Emotion/Reward &
Punishment on EF
• Activation in left DLPFC (yellow) when
viewing cues signaling the potential for
reward (associated with faster RTs)
• Activation in right OFC (green) when
receiving rewards (associated with faster
RT)
•
R
z = 28
Activation in left OFC (red) when
receiving punishments (also associated
with faster RT)
R
z = -8
Influence of Emotion/Reward &
Punishment on EF
• Preliminary results thus:
– Confirm effects of both affective and motivational
contexts on executive function
– Replicate opposing patterns of lateralization for
DLPFC and OFC
– Allow us to examine timing of regional activity,
connectivity, relationships of regional and
temporal dynamics to emotional disposition
– Can be extended to examine dysfunctional
relationships in depression & anxiety
Effects of Anxiety on Selection Among
Competing Options
Hannah R. Snyder & Yuko Munakata (Project 5) in collaboration with:
Marie T. Banich (Project 1)
Tim Curran & Erika Nyhus (Imaging Core)
With consultation from Project 3
Anxiety and Uncertainty
• Anxious apprehension (worry) is linked to
intolerance of uncertainty (e.g. Ladouceur, Talbot & Dugas, 1997),
decision-making problems, and indecisiveness
(e.g. Sachdev & Malhi, 2005).
• Prominent symptoms of anxiety disorders
including GAD and OCD.
• Why?
• Approach this question using our framework
for understanding one aspect of EF: selection
among competing options.
Selection Among Competing Options
•
•
We constantly face the need to choose one option from among
multiple valid choices.
• e.g. grocery shopping, selecting a retirement plan, or choosing a
word to express a thought.
Selecting between multiple options is time-consuming and effortful
(e.g. Iyenger & Lepper, 2000; Sethi-Iyenger et al., 2004; Snyder & Munakata, 2008).
•
Left ventrolateral prefrontal cortex (VLPFC) is involved in selection (e.g.
Thompson-Schill et al., 1997,1998; Schnur et al., 2009).
• Illinois center colleagues Engels, Heller, & Miller have shown that left
•
VLPFC is involved in anxious apprehension.
What specific mechanisms might support selection, and how might
they be affected by anxiety?
Selection Among Competing Options
• Test neural network predictions about selection
using a well-controlled language-production task:
verb generation.
Neural Network Model
• Demonstrates that competitive, inhibitory dynamics among
neurons in prefrontal cortical networks support selection
among competing alternatives.
– Amplify activity of most active representation and
suppress activity of competing representations, via
inhibitory, GABAergic interneurons.
Pyramidal
Cell
+
+
GABAergic
Interneurons
Pyramidal
Cell
Neural Network Model
• Suggest that prefrontal GABA function plays key role in
selection and breakdown of this process.
• Makes sense of findings which were previously
disconnected from each other, linking anxiety to:
• Reduced GABA (e.g. Kalueff & Nutt., 2007)
• VLPFC dysfunction (e.g. Engles et al., 2007).
Competitive Inhibition & Selection
• Neural network predictions:
– Anxiety (reduced neural inhibition) impairs selection and
associated VLPFC activity, even in a simple, non-affective
language-production task.
– The GABA agonist midazolam (increased neural
inhibition) improves selection.
– Retrieval from semantic memory is unaffected.
• These predictions were supported in 3 studies.
Anxiety:
Decreased Inhibition Impairs Selection
Network Predictions
•Reduced competitive
inhibition in the VLPFC layer
impairs selection.
39
Participants (RTs)
•High anxious apprehension
participants have impaired
selection.
•No effect on retrieval.
Anxiety: Decreased Inhibition Impairs
VLPFC Function During Selection
• Anxious apprehension
correlates negatively
VLPFC ROI
with VLPFC activity
during selection (when
retrieval demands are
low).
• No correlation during
retrieval.
Midazolam:
Increased Inhibition Improves Selection
Network Predictions
•Increased competitive
inhibition in the VLPFC layer
improves selection when
retrieval demands are low.
41
Participants (RTs)
•Midazolam improves
selection when retrieval
demands are low.
•No effect on retrieval.
Conclusions
•Neural network model suggests that competitive inhibitory
dynamics in prefrontal networks are critical for selection.
•As predicted by model, participants high in anxious
apprehension (linked to reduced GABAergic function) show
impaired selection but not retrieval.
- Consistent with clinical evidence for decision-making
problems and intolerance of uncertainty in anxiety disorders.
•Participants high in anxious apprehension show reduced left
VLPFC recruitment during selection.
-Could represent failure to activate inhibitory interneurons.
Conclusions (cont.)
•As predicted by model, midazolam (GABA agonist) improves
selection when retrieval demands are low.
-Suggests GABA agonists may be beneficial in treating
cognitive, in addition to affective, symptoms of anxiety
disorders.
Ongoing and Future Directions
•Study with selected high and low anxiety participants
across multiple selection tasks.
•Comparing underdetermined to prepotent
competition (behavioral and fMRI studies).
•Effects of depression on controlled retrieval.
Thanks!
•Professional research assistants: Paula Villar,
Kirsten Orcutt, and Luka Ruzic
•Undergraduate honors thesis students: Natalie
Hutchison and Teesa Dutta
•Clinical collaborators: Rosi Kaiser and Mark
Whisman
•All DEFD members for helpful input.
Major Component Processes
Involved in Executive Function
Friedman, Hewitt, Willcutt, Young, Smolen, Miyake,
O’Reilly, Hazy, Herd, Brant, Chatham
46
Three Components of EFs
• Inhibition
– Stopping prepotent (dominant or automatic)
responses (e.g., stop-signal)
• Updating
– Monitoring and rapid addition/deletion of the
contents of working memory (e.g., n-back)
• Shifting
– Switching flexibly between tasks or mental sets
(e.g., number-letter)
47
Unity and Diversity
Plus-Minus
Number-Letter
.59
.57
Shifting
.46
Local-Global
Keep Track
Tone Monitoring
.56
.46
.45
.63
Updating
.42
Letter Memory
Stroop
Stop Signal
.63
.40
.33
Inhibition
.57
Antisaccade
Miyake et al. (2000), Cognitive Psychology
48
Unity and Diversity of EFs
Unity
Updating
Ability
=
Shifting
Ability
=
Inhibition
Ability
=
Common EF
Diversity
+
Updating-Specific
+
Shifting-Specific
+
Inhibition-Specific
49
Nested Factor Model
Common
EF
.46
Anti
.58
Stop
.43
Stroop
.41
Keep
Updating
specific
Shifting
specific
.54 .53 .22
.49 .46 .58
.44
.37
Letter
.47
S2ba
.42
Num
.46
Col
Cat
50
Twin Study of EFs
• Colorado Longitudinal Twin Study (LTS)
• 159 MZ (identical twin) pairs & 134 DZ (fraternal
twin) pairs
• 9 EF tasks to construct latent variables
• Compare MZ and DZ twin data to estimate:
– A: Additive genetic (heritability)
– C: Shared environment
– E: Nonshared environment
51
Genetic Unity and Diversity
98% 0%
A
C
2%
100% 0% 0%
A
E
Common
EF
Anti
Stop
C
76% 0% 24%
A
E
Updating
specific
Stroop
Keep
Letter
C
E
Shifting
specific
S2ba
Num
Col
Cat
Friedman et al. (2008), Journal of Experimental Psychology: General
52
Translational Implications
• Components show different relations to a
range of behavioral and psychological
problems:
•
•
•
•
•
Depression
Behavioral disinhibition
Attention problems
Early (toddler-age) self-restraint
Sleep problems
• More precision about EF profiles
53
Biological Basis of Unity and
Diversity
• Emerges from involvement of multiple brain areas
• Different brain areas suited to different operations
– PFC for active maintenance
– Basal ganglia for updating PFC
• Different influences of genes in these areas
– COMT in PFC (Val158Met in COMT gene)
– D2 receptors in striatum (C957T in DRD2 gene)
54
Example Model: N-Back
Hidden
Stimulus
& Parietal Input
Ventral Striatum (PVLV)
Prefrontal Cortex
Verbal &
Manual Output
Dorsal Striatum (Matrix & SNr)
Inputs: serial order & item information. Outputs: verbal & manual output
Leabra framework PBWM architecture
(O’Reilly, 2001)
(Hazy, Frank & O’Reilly, 2006)
55
Example Model: N-Back
Prefrontal
Maintains information
with intrinsic & recurrent
maintenance currents
Striatal Matrix
Decides when to maintain info
in PFC; trained with RL on
predicted reward (PVLV)
56
Modeling Genetic Influences
• Use a number of polymorphisms known to
affect DA.
– e.g.
– COMT val/met → affects levels of tonic DA in
PFC
– DRD2 TAQ1A SNP → affects density of D2
receptors in striatum
• Simulate those effects within model
57
Example Manipulation: COMT
• COMT: val/met polymorphism
• COMT removes DA in PFC
• Met/met have higher tonic DA vs. relatively
low (val/val) or middle (val/met) levels
• Met/met individuals perform better on a
range of cognitive tasks (Savitz et al., 2006)
58
Modeling COMT effects
• Met variant
– increased DA in PFC
– excites active neurons, inhibits less active
– enhances recurrent NMDA channels effects
– Thought to increase signal-to-noise ratio in
PFC
• Modeled as increased gain of PFC neurons
59
val/met COMT/mid dopamine/mid gain in PFC
01/05/10
60
val/val COMT/low dopamine/low gain in PFC
01/05/10
61
met/met COMT/high dopamine/high gain in PFC
01/05/10
62
Gain curves for
PFC neurons
01/05/10
63
PFC Gain Affects Performance
• Gain manipulations replicate observed inverse
U-shaped curve for DA effects
• COMT polymorphisms plus amphetamine (e.g.,
Mattay et al., 2003)
01/05/10
64
Models Test the Simple Story
Goal maintenance
(PFC)
Specificity of Gating
(BG)
Slipperiness of Reps
(PFC)
Common
EF
Updating
specific
Shifting
specific
Anti
Stop
Stroop
Keep
Letter
S2ba
Num
Col
Cat
Modeling can reveal nonlinear effects,
interactions between systems, and divisions
of labor over learning
65
Planned Work

Model the rest of the tasks:





Inhibition: Stroop, antisaccade, stop signal
Updating: Keep Track, n-Back, Letter Memory
Shifting: color-shape, letter-number, vowel-cons
Unify models
Predict (and explain) effects of specific genes on
components

Gene effects on brain measures:
activation by area (BOLD), latency by area (ERP)
66
Major component processes involved in
executive function:
Assessment of Executive Function Components
Stacie Warren, with Wendy Heller and Gregory A. Miller, post-doc
Dave Towers, and Center colleagues Marie Banich, Naomi
Friedman, Akira Miyake
Psychology Department, University of Illinois at UrbanaChampaign
Framework for Test Selection
• Goals
– Target 3 EF domains: shifting, inhibition, and updating
• Sensitivity
– Detect effects of personality (e.g., positive/negative trait
affect) and psychopathology (depression/anxiety) on EF
– Selective enough to engage prefrontal regions such as
DLPFC
– Level of difficulty
• Floor/ceiling effects
• Task Simplicity
– Isolate EF components we are interested in
• Task impurity problem
Framework for Test Selection
• Comparable nonverbal analogues to verbal tasks
• Multiple measures for each domain
– Helps alleviate task impurity problem & low reliability
• Tolerability and practicality
– Longer the battery the more reliable
– Increases likelihood for boredom, dropping out
• IQ and Processing Speed measures
– Differential deficit
Approach to Test Selection
• Comprehensive sampling of EF performance
• Tasks identified as critically dependent on one of the
subcomponents of EF?
– “Executive Function” tasks
• Empirically supported EF component tasks
– Miyake, Friedman, and colleagues
• Clinical Measures
– D-KEFS (Delis-Kaplan Executive Functioning System)
D-KEFS
• A relatively new measure that attempts to isolate
component processes necessary for EF task
performance.
• Consists of tests that are adaptations of tests
currently used for assessing EF
• Greatly improved on earlier versions of these tasks
by providing process scores that offer insight into
performance scores
• Normed on a sample of 1,700 across US, ages 8-89
EF Tasks
• Response Inhibition
RED BLUE GREEN YELLOW
– DKEFS Stroop
– Stop Signal Task
– TOL (updated computerized version)
• Switching
– DKEFS
• Trails, Category Fluency, Design Fluency, Stroop
– Plus-Minus
EF Tasks
• Updating
– Keep Track
– Letter Memory
– Spatial Updating Task (Heller/Miller lab
developed)
• Visuospatial updating task
• More details in a bit
Additional Tasks
• Processing Speed
– WAIS Coding & Symbol Search
• IQ
– WTAR (VIQ)
– WAIS Block Design
• PASAT-100
– Attentional control, divided attention, working memory
• Subjective Reports of EF in Everyday Life
– Behavior Rating Inventory of Executive Function (BRIEF):
self and informant reports
Spatial Updating Task
• Why?
– Lack of visuospatial tasks that target updating
• WMC, dual-task components, too many operations, etc.
– “Gold standard” is n-back
• Requires significant attentional control
– Spatial task without verbal tags
• Assisted n-back
• Demo task
135 degrees
90 degrees
45 degrees
180 degrees
0.1
0.2
0 / 360 degrees
0.3
225 degrees
270 degrees
0.4
0.5
0.6
0.7
0.8
0.9
315 degrees
Spatial Updating Task
• How?
– Used Letter Memory as a template
– Matlab randomly generated box locations
• Circular grid used to avoid verbal tags & reduce effects
of saccades
– “Real” trial sequence lengths of 9, 11, & 13
• Randomly generated targets within a sequence length
• Avoided recognizable spatial patterns
135 degrees
90 degrees
45 degrees
180 degrees
0.1
0.2
0 / 360 degrees
0.3
225 degrees
270 degrees
0.4
0.5
0.6
0.7
0.8
0.9
315 degrees
Task Piloting
• What are we measuring?
– Errors within a sequence
– Time
• Time it takes to respond from response cue (“???”) to
first mouse click
• Total time it takes to respond within a step
– Distance
– Velocity
Pilot Data
• First two pilot rounds N=19 (informal, lab
members, friends)
– 3 vs. 4 back
– Some sequences revised
• Third round of piloting
• N=13; 18-20 years, 11 female
• Reliability: .96
Average Error*
0.3
0.2
0.1
0.0
8
9
10
11
Trial Block Length
12
13
*Total commission errors averaged across subjects; 1 diamond per trial type
14
0.5
Average Error*
0.4
0.3
0.2
0.1
0.0
9
11
Trial Block Length
*Total commission errors averaged across trial type; 1 line per subject
13
The Nature of Inhibitory Processes:
Is stopping or monitoring the crucial executive
component to inhibitory control?
Chris Chatham & Yuko Munakata
(Project 5) in collaboration with
Marie Banich, Tim Curran, Albert Kim
83
Fractionating Inhibitory
Control
• Inhibitory control requires multiple
subprocesses. Among them:
“Context
Monitoring”
1. Vigilance
2. Detection of the need for stopping or
suppression, often as cued by infrequent or
unusual stimuli
3. Stopping and/or suppression
• Most theories emphasize #3;
–
but Context-monitoring may account for
some of the variance thought to be
explained by #3
•
E.g., the involvement of the right inferior
frontal gyrus in inhibitory control
84
Empirical Approach
(Chatham, Claus & Munakata, in prep; Chatham, Banich, Curran, Kim & Munakata, in
prep)
Task Stimuli:
No Signal; 75% of trials
200 ms
time
Stop Signal or Oddball; 25% of trials
Rest of TR
200 ms
X = {100, 150,
250, 300} ms
Rest of TR
time
2x2 Task Design:
Fixations both intermixed & blocked: a hybrid fMRI design
85
Unique predictions of the
Context Monitoring Account
• Temporal dynamics:
– monitoring predicts both sustained and transient
components
• Same parts of rIFG should be active in both tasks
– despite their different stopping demands
• rIFG may be more active in oddball task
– Oddball task presented first
– Thus signal stimulus is most unusual/infrequent then
• Also: individual differences, pupillometry
86
Context monitoring better accounts for
BOLD in RIFG than stopping
n=18, thresholded at 2.58
Task > Fixation
(sustained act)
Signal Trials > No Signal Trials
(transient act)
Blue: Stop Task
Red: Oddball Task
Blue: Stop Task
Red: Oddball Task
Signal Trials > No Signal Trials
(transient act)
Blue: Stop > Oddball (empty map)
Red: Oddball > Stop (cluster in rIFG)
Similar results achieved w/ ERP:
a shared principal component
above the right frontal lobe
(a second sample of 38 subjects)
Oddball task
Stop task
87
The Nature of Inhibitory Processes:
Monitoring may be the crucial executive
component to inhibitory control
• BOLD & ERPs in rIFG
– do not show unique patterns in a task that demands stopping,
relative to one that only demands context monitoring
• In fact, rIFG is more strongly recruited by the latter task
• Individual differences… (in a third sample of 96 subjects)
– are not uniquely captured by a task that demands stopping,
relative to one that only demands context monitoring
• In fact, more variance is explained by the latter task
• Temporal dynamics of rIFG
– Have both sustained and transient components, consistent with
a context monitoring function
– Time course of activity is highly similar across tasks (ERP
temporal PCA)
88
Future Directions
• ROI analyses (44 vs 45 vs 47)
• Functional connectivity/PPI
• Neural network modeling (w/ Project 2)
89
The Nature of Inhibitory
Processing
Determinants of Executive Function and
Dysfunction
B. Depue, M. Banich, K. Mackiewicz, G. Burgess, T.
Curran, R. O’Reilly, Y. Munakata, C. Chatham, H.
Snyder
90
Current Implications
• Our studies examining inhibitory function
have suggested:
– That areas of right LPFC appear to be involved in
inhibitory control across multiple domains
• Motor (well studied)
• Memory/Thought
• Emotional
– Inhibitory control appears to down-regulate
cortices that support representations of material
involved in the specific task at hand
91
Think/No-think Task
• Do inhibitory mechanisms act on pictorial and
emotional memory representations?
• Three phases:
– Training
– Experimental
– Testing
92
Blocked Condition
Training Phase: 40 Negative Pairs
+
93
Training Phase:
Practice Until Recognition >95%
or
+
or
94
Most Importantly
• From this point on, no external representation of the
target is shown
• Individuals can only manipulate the internal
components of memory representation
95
Experimental Phase: 240 Trials
Think
+
Cue
No-Think
Think of
previous
associated
picture
Target
+
Cue
Do not let
previous
associated
picture enter
consciousness
Target
96
Repetition Manipulation
0 “Baseline”
12x
Training
Training
+
+
Experimental
Experimental
Randomly
distributed
Testing
+ Short desc.
+
+
+
+
+
+
+
+
+
+
+
+
Testing
+ Short desc.
97
Testing Phase: Cued Recall
description
+Short
_____
Cue
Target
description
+Short
_____
Cue
Target
98
Behavioral Results - TNT
Percentage Recalled
75
70
m (Base) = 62.5 %
m (T)
= 71.1 %
m (NT)
= 53.2 %
65
60
T
NT
55
50
0
12
Repetitions
99
Imaging Results - TNT
• Sources of cognitive control/inhibitory control
• Sites of where that control is directed
100
Sources of Cognitive Control
y=22
z=29
z=3
rMFG
rSFG
rIFG
rMFG
rIFG
NT>T
• Right PFC
Involved in executive functions/cognitive control
Increased activity for NT trials suggests rLPFC
increased involvement during inhibition
101
1. Sites of Cognitive Control
z=5
y=-57
Pul
y=-90
z=-16
BA18
FG
BA17
BA17
NT>T
• Pulvinar
 Controlling the flow of visual information to cortex
• Visual cortex
 Visual areas and fusiform gyrus
• Known to process visual representations, selective for objects/faces
102
2. Memory Processes and
Emotional Components
y=-22
y=-14
Hip
y=1
Hip
Hip
y=5
Hip Amy Amy Hip
Amy Amy
NT>T
• Hippocampus/Parahippocampal Gyrus
 Highly involved in encoding, consolidation, and retrieval
 Binds associative components of episodic/semantic memory
• Amygdala
 Responsible for generating emotional responses
 Bidirectional connectivity for modulation of learning and memory
103
Important !
• Looking at signal change analyses shows decreased
activity below baseline for NT trials in
 Sensory cortex (Pulvinar, Fusiform gyrus)
 Emotion and Memory (Hippocampus, Amygdala)
• People appear to inhibit/down-regulate brain areas
underlying sensory gating, sensory representation,
emotional components and memory processes of
memory representation
104
Functional Connectivity Analysis
• Looking at the functional connections of brain
regions over time
• Examining NT trials>baseline
• Two networks were identified:
– rIFG functionally connected with the Pulvinar and
Fusiform Gyrus
– rMFG functionally connected with the
Hippocampus and Amygdala
105
Functional Connectivity Analysis
IFG&Pul
IFG&FG
MFG-Hip
MFG&Amy
Correlation Coefficient
0.2
*
+
#
0
-0.2
-0.4
+
+
-0.6
-0.8
+
+
#
#
-1
= p<.05
1
= p<.01
= p<.001
2
3
4
Quartile
106
Summary - TNT
•
Components of visual information in No-Think trials
appear to be inhibited

•
This mechanism is invoked from earliest attempts at inhibition
Inhibition also involves decreasing activity in
regions involved in memory and emotion

These mechanisms appear to require repeated attempts at inhibition
107
Thought Suppression
• To examine whether we get similar right LPFC
regions involved in the control over memories
as found with TNT
• To determine whether the activity of these
regions are specific to the inhibition of
thoughts or the manipulation of thoughts
more generally
108
Paradigm
• 32 total stimuli
– 8 neutral color pictures (e.g., peacock)
– 8 neutral black & white pictures (e.g., penguin)
– 16 neutral melodies with words (e.g, happy birthday)
• 4 conditions
–
–
–
–
Maintain
Switch
Suppress
Clear
109
Paradigm
Maintain
+
Fixation
2 – 16 seconds
Maintain
Image
4 seconds
Cognitive
Manipulation
4 seconds
Switch
+
Fixation
2 – 16 seconds
Zebra
Image
4 seconds
Cognitive
Manipulation
4 seconds
+
…
Fixation
2 – 16 seconds
110
Thought Suppression
Pattern of activation in the occipital cortex and
other visual regions suggests that participants are
complying with the task demands
Maintain>+++++
Maintain>Switch
Maintain>Clear
Maintain>Suppress
X=48
Z=-20
111
Linear Regression
• Main > Switch > Clear > Suppress (visual areas)
• Linear pattern suggests that there is increased
right LPFC as a representation or supporting
cortices must be inhibited/manipulated
Z=-20
X=48
112
Future Studies
• ERP and inhibition over memory retrieval
(connection with Imaging Core)
• Examining the TNT with PTSD (connection
with Project 3)
113
ERP and Memory Inhibition
114
ERP and Memory Inhibition
• Parietal areas show differential processing for NT and
T items
– Such that NT items show reduced or possible blocked
retrieval
• Continue to exam results with source localization and
seeding regions with fMRI data
115
PTSD and Memory Inhibition
• Collaboration with Denver VA
• Examining the integrity of structure/volume of
hippocampus
• Examining the feasibility of using the TNT with
war veterans
116
The influence of learning & development on executive
function:
Temporal dynamics in
cognitive control
Chris Chatham & Yuko Munakata
(Project 5) in collaboration with
Michael Frank
117
A Large Developmental Change in EF:
The temporal dynamics of control
• Age-related change in EF widely thought to
reflect changes in the speed or strength of EFs
– in goal maintenance or active inhibition of
irrelevant information
– Original DEFD studies built on this assumption of
quantitative change
• We have evidence for a more drastic
qualitative shift:
– age-related change in when control is engaged
(Chatham, Frank, & Munakata, 2009; Chatham & Munakata, in progress)
118
Cog Control Dynamics in AX-CPT
• Adults maintain the informative context info provided by cues
•
•
“A” predicts target (87.5% of the time)
• Might encourage AY errors!
“B” predicts nontarget (100% of the time!)
• Might reduce BX errors!
• Expected: some modulation of children’s behavior due to the
maintained context
•
Observed: No maintained context; retrieval only when necessary
• RT slowing, individual differences, sequence effects, speed-accuracy
tradeoffs, pupillometry
119
Two Examples of Reactive -> Proactive Transition
(Chatham, Frank, & Munakata, 2009)
Mental Effort
(pupil diameter):
•Probe-period effort
in 3.5 year olds
•Delay-period effort
in 8-year-olds
Individual Differences in
RT:
•Probe-Driven in 3-yrolds,
•Cue-Driven in 8-yr-olds
120
When does the reactive -> proactive transition occur?
Around 5 years of age. (Chatham & Munakata, in progress)
Cue
• 6 year olds:
Delay
Probe
• 5 year olds:
121
The Influence of Distraction on
Proactive Control
• Delay-period distractors disrupt proactive control (as
in 6-year-olds)
• Distractors have less effect on reactive control (as in
5-year-olds)
(higher values = more proactive)
Working Memory Index
0.9
0.8
Distractors
0.7
No Distractors
0.6
0.5
0.4
0.3
0.2
0.1
0
5-year-olds
6-year-olds
122
Learning & Development of EFs
• Development involves not only the strengthening
of EFs, but also a change in how they are
deployed:
Reactive - EFs engaged only as needed in the moment
Proactive - EFs engaged to meet an anticipated demand
• The reactive to proactive shift may be graded
– e.g., 5-year-olds have both cue & probe-driven
relationships in RT
123
Future Directions
• What advantages might a reactive mode
confer to learning?
– Neural network modeling
• How task-dependent is the use of reactive and
proactive mechanisms?
– Convergent measures of reactive control
• What are the neural correlates of reactive
control?
124
Developmental Differences in Toddler’s
Behavioral Restraint Predict Executive Control
Abilities 14 Years Later
Naomi P. Friedman, Akira Miyake, & John Hewitt
University of Colorado at Boulder
Self-Regulation and Executive Functions
• Individual differences in lab-based EF tasks can
capture variation in self-regulation
– EF abilities are substantially related to:
• Attention problems at school during adolescence
(Friedman et al., 2007, Psychological Science)
• Externalizing behavior problems in late
adolescence (Young et al., 2009, Journal of Abnormal
Psychology)
• Emergence of self-regulatory abilities?
Self-Regulation in Early Childhood &
Beyond
• Systematic variation in behavioral restraint exists
in early childhood
– Delay of gratification (Michel’s work)
– Prohibition (Kochanska’s work)
• It is developmentally stable and is predictive of
success later in life
– Academic achievement and social functioning
(Duckworth & Seligman, 2006; Michel, Shoda, & Peake, 1988; Shoda,
Michel, & Peake, 1990)
Main Questions for the Study
 Are individual differences in behavioral selfrestraint during early childhood related to
individual differences in EF abilities observed later
in life?
 If so, which aspects of EF abilities are most closely
related to early self-restraint?
 To what extent is the longitudinal relationship
genetically mediated?
Unity and Diversity of EFs
Keep Track
.65
Letter Memory
.66
Spatial 2-Back
.46
Number-Letter
.66
Color-Shape
.63
Category Switch
.74
Stroop
.42
Stop Signal
.53
Antisaccade
Updating
.40
.74
Shifting
.73
Inhibition
.44
Friedman et al. (2008) JEP:General, N = 582 (Longitudinal Twin Sample)
Unity and Diversity of EFs
Unity
Updating
Ability
=
Shifting
Ability
=
Inhibition
Ability
=
Common EF
Diversity
+
UpdatingSpecific
+
ShiftingSpecific
+
InhibitionSpecific
Active maintenance
of goals and goalrelated information?
Main Questions for the Study
• Are individual differences in behavioral selfrestraint during early childhood related to
individual differences in EF abilities observed
later in life?
• If so, which aspects of EF abilities are most closely
related to early self-restraint?
• To what extent is the longitudinal relationship
genetically mediated?
Unity and Diversity of Genetic Influences
Unity
Updating
Ability
=
Shifting
Ability
=
Inhibition
Ability
=
Diversity
+
Common EF
A
98%
C
0%
+
Updating-Specific
A
100
%
C
0%
A
76%
C
0%
E
0%
Shifting-Specific
E
24%
E
2%
Friedman et al. (2008) JEP:General, N = 582 (Longitudinal Twin Sample)
The Sample and Tasks
• 822 individual twins from the Colorado
Longitudinal Twin Study sample
– All from same-sex twin pairs raised together
– Normally distributed IQ
• Task administration
– Prohibition task: Ages 14, 20, 24, & 36 months
– WAIS IQ: Age 16
– EF test battery: Age 17
Prohibition Task
• Procedure:
– The experimenter draws attention to an attractive
toy (a glitter wand)
– “Now, don’t touch”
• Dependent measure:
– Whether the child touched the toy within 30 s
Boy A (24 months old)
Boy B (24 months old)
Latent class growth modeling identified 2
distinct groups of children
Group 1
Group 2
Group Differences in EF (Age 17)
Unity
Diversity
=
+
Shifting
Ability
=
Common EF
+
Inhibition
Ability
=
.45 SDs above
Updating
Ability
Updating-Specific
.02 SDs below
Shifting-Specific
.34 SDs below
WAIS IQ
.24 SDs above
Genetic and Environmental Correlations
Unity
Updating
Ability
=
Shifting
Ability
=
Inhibition
Ability
Diversity
+
Updating-Specific
A
98%
Common EF
+
Shifting-Specific
A
80%
=
A
94%
C
0%
E
6%
.54*
E
2%
C
0%
−.03
E
20%
C
0%
−.66*
A
C
.29
E
28% 32% 40%
Group Membership
Summary of the Main Results
 Developmental differences in toddler’s
behavioral self-restraint predict EF Abilities in
early adulthood
 Early prohibition performance is related:
 positively to Common EF
 negatively to Shifting-Specific
 This longitudinal relationship is due to common
genetic influences:
 Common EF = .54
 Shifting-Specific = −.66
Another Example of Opposing Effects Observed for
Common EF and Shifting-Specific Factors
Unity
Updating
Ability
=
Shifting
Ability
=
Inhibition
Ability
=
Diversity
+
Common EF
A
99%
C
0%
E
1%
.57
+
Updating-Specific
A
100
%
C
0%
A
78%
C
0%
E
0%
Shifting-Specific
E
22%
−.20
.56
−.05
C
A
E
75% 10% 15%
WAIS IQ
Discussion and Conclusion
• Shifting ability (measured as switch costs) may
better be viewed as a mixture of two opposing
forces
• Common EF = stability (goal maintenance)
• Shifting-specific = flexibility
• Early behavioral restraint is a precursor of
later executive functioning
• Genetic factors contribute in part to this
developmental stability