Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Cognitive Neuroscience of Thought
2006 Summer School, Sofia, Bulgaria
July 4, 2006
Kalina Christoff
Goal-directed Thought
The Contribution of Lateral Prefrontal Cortex
How do we think about goal-directed thought?
sfg
10
A set of distinct component processes
9 9/46
46 mfg
45
47/12
ifg
• can be described fairly precisely
Îpsychological
Î
nature
neural basis
• are organized by particular principles
RLPFC
(BA 10)
DLPFC
(BA 9, 46, 9/46)
VLPFC
(BA 45, 47/12)
Relational Complexity Framework
(Halford et al., 1984)
Rostrolateral prefrontal cortex
Dorsolateral prefrontal cortex
Ventrolateral prefrontal cortex
The ability of relational integration
Occurs late in child development (Halford, 1984)
Level 0:
0 relations must be considered
Level 1:
1 relation must be considered
Unique to humans (Tomasello & Call, 1997)
Linked to prefrontal cortex (Robin & Holyoak, 1995)
Level 2:
2 relations must be considered
relational integration
2006 Summer School, Sofia, Bulgaria
Specific deficit with frontal lesions (Waltz et al., 1999)
1
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Raven’s Progressive Matrices
0-relational
Raven’s Progressive Matrices
1-relational
Raven’s Progressive Matrices
Event-related Design
2-relational
fMRI acquisition
Event-related Design
Parameters
1.5 T GE Signa scanner
TR = 2 s
16 axial slices
7 mm slice thickness
no spacing between slices
spiral pulse sequence
bite-bar
10 participants
Functional analysis
SPM 99b
random effects model
2006 Summer School, Sofia, Bulgaria
2
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Behavioral Performance
Mean Response Time (RT)
Mean Accuracy
100%
99.6%
Two Different Sources of
Activation
5.66
6
96.7%
85.0%
RT (sec)
Percent Accurate
5
4.19
1-rel
vs. Proc2
Proc3
Condition A
Proc1
Proc2
Proc3
Condition B
Proc1
vs. Proc2
Proc3
Condition B
A vs. B
2.78
3
2
0
0-rel
Condition A
Proc1
Proc2
Proc3
4
1
0%
Novel
processes
2-rel
0-rel
1-rel
2-rel
Shared
processes
Problem Type
Problem Type
(Christoff et al., 2001)
Study 1: Relational Integration during Reasoning
(Christoff et al., 2001)
Relational Integration during Reasoning
Modeling the Hemodynamic Response
Fixed HRF
Traditional modeling approach:
1 rel > 0 rel
Fixed HRF
Regressor functions based on the canonical HRF
2 rel > 1 rel
(Christoff et al., 2001)
Study 1: Relational Integration during Reasoning
RT-convolved HRF
Modeling the Hemodynamic Response
An alternative approach:
RT-Convolved HRF
1 rel > 0 rel
RT-based box-car function
2.36 s
3.29 s
4.91 s
4.17 s
2.79 s
3.85 s
2 rel > 1 rel
Regressor function after convolving
Study 1: Relational Integration during Reasoning
2006 Summer School, Sofia, Bulgaria
(Christoff et al., 2001)
3
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Anatomical Regions of Interest
Fixed HRF
2 rel > 1 rel
2 rel > 1 rel
RT-convolved HRF
All Trials
RT-Matched Trials Only
RLPFC
(BA10)
RLPFC
(BA10)
DLPFC
(BA46)
Trials matched for response time and accuracy
R
L
N problems
Mean RT (±S.D.)
36
4.05 sec (±0.68)
0-rel
z=4
DLPFC
(BA9)
Average time course
1-rel
36
4.07 sec (±0.70)
2-rel
36
4.07 sec (±0.66)
R
L
0-relational
1-relational
(Christoff et al., 2001)
Study 1: Relational Integration during Reasoning
z = 32
2-relational
(Christoff et al., 2001)
Study 1: Relational Integration during Reasoning
RT-Matched Trials Only
MFG
BA10
RT-Matched Trials Only
MFG
BA10
MFG
BA10
R
L
BA46
R
L
z=4
z=4
MFG
BA9
Average time course
R
L
0-relational
1-relational
2-relational
Average time course
R
L
0-relational
1-relational
z = 32
2006 Summer School, Sofia, Bulgaria
2-relational
z = 32
4
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Relational integration
recruited RLPFC & DLPFC regions
RLPFC
Î
(Area 10)
RLPFC
greatest specificity
Î
Activated across wide range of domains
• reasoning and problem solving
• working memory
• memory retrieval
• attention
(Area 10)
RLPFC
RLPFC
Î
(Area 10)
Î
(Area 10)
RLPFC
Domain-independent functions
(BA 10)
• working with the products of ones own thought
processes
• introspective thought
• meta-cognitive representation
DLPFC
Activated during tasks
at the highest level of
cognitive complexity
(BA 9, 46, 9/46)
VLPFC
(BA 45, 47/12)
RLPFC
Is its recruitment
always hierarchical?
Î
(Area 10)
Rest vs Arrows
RLPFC
(BA 10)
Mental processes
during rest
DLPFC
(BA 9, 46, 9/46)
VLPFC
(BA 45, 47/12)
2006 Summer School, Sofia, Bulgaria
5
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Different levels of abstraction in thinking
Executive processes:
abstract vs. concrete information
Thought processes performed upon information that is
RLPFC
• Computational models of prefrontal function
(O’Reilly et al., 2002)
Highly abstract
(BA 10)
DLPFC
Relatively abstract
VLPFC
Relatively concrete
(BA 9, 46, 9/46)
(BA 45, 47/12)
• Developmental models of executive functions
(Zelazo et al., 1999, 2004)
Patterns of connectivity
Continuum
Pandya & Barnes (1987)
Abstract
Concrete
Redness
Red Car
Red BMW
Your own
BMW
Rostrocaudal (anterior-to-posterior) gradient:
More anterior PFC regions are connected to areas of
increasing order of association in posterior cortices.
Solving Anagrams
Abstract
Abstract
200
250
300
Medium
350
400
450
500
Concrete
550
600
650
APealp
Medium
Concrete
Dncae
Dkes
700
Concreteness ratings for nouns
Hmar
Tpir
F o do
Paivio et al. (1968), MRC Psycholinguistic Database (Wilson, 1988)
Gacre
Sngo
BOltet
Christoff, Keramatian et al. (in prep)
2006 Summer School, Sofia, Bulgaria
6
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
Solving Anagrams
Abstract
Medium
Abstract
Medium
Concrete
Instructions
(2 sec)
Concrete
APealp
Dncae
Dkes
(Appeal)
(Dance)
(Desk)
Stimuli
(4 sec each)
Hmar
Tpir
F o do
(Harm)
(Trip)
(Food)
Abstract
Medium
Concrete
Whis
Fvere
Bkna
(wish)
(fever)
(bank)
8 stimuli per block, 12 blocks per condition
Gacre
Sngo
BOltet
(Grace)
(Song)
(Bottle)
conditions matched for
• word frequency (M=53)
• word length (M=5.37)
• number of syllables (M=1.5)
• difficulty
fMRI acquisition
Abstractness or Complexity
Level of abstraction
Level of difficulty
Volume
Slices
Reaction time
Level of difficulty
Number of errors
Behavioral Pilot Study
Dacen
Danec
Dance
Very
Fairly
Easy
Solution
difficult
difficult
Mean RT (ms)
Decan
Accuracy
Mean Accuracy (%)
Reaction Times
Abstract
1629.7 ms
Medium Concrete
1667.2 ms 1667.7 ms
Abstract
60.5%
Medium
59.1%
Concrete
60.3%
Christoff, Keramatian et al. (in prep)
2006 Summer School, Sofia, Bulgaria
7
Cognitive Neuroscience of Thought
Kalina Christoff, UBC
July 4, 2006
sfg
Abstract
Medium
Concrete
[abs vs. (con+med)/2]
[med vs. (abs+con)/2]
[con vs. (abs+med)/2]
10
9 9/46
46 mfg
45
47/12
ifg
RLPFC
(BA 10)
DLPFC
(BA 9, 46, 9/46)
RLPFC
(BA10/46)
DLPFC
(BA 9/46)
VLPFC
VLPFC
(BA 47/11)
(BA 45, 47/12)
Hierarchical
Complexity dependent
Parallel
Levels of abstraction
Christoff, Keramatian et al. (in prep)
2006 Summer School, Sofia, Bulgaria
8