testing the biased competition model of attention in the ...representing abstract information. •...

1
P<0.05 (corrected, cluster level) IFG insula SMG IPS R L When there was competition between abstract rules, as in the incongruent trials, participants were slower and less accurate, and this effect was accompanied by increased activity in areas associated with cognitive control. Right IFG serves as the source of top-down bias in response to competition between task rules by enhancing activity in brain areas representing relevant information, rather than by suppressing activity in brain areas representing irrelevant information. The neural pattern of right IFG contains rule information, which provides corroborating evidence that right IFG mediates competition between brain areas representing abstract information. The results suggest that “biased competition” serves as a mechanism to guide the selection of task-relevant abstract information. Testing the Biased Competition Model of Attention in the Selection of Abstract Task Rules Yi-Shin Sheu 1 , Susan Courtney 1,2,3 1 Department of Psychological & Brain Sciences, Johns Hopkins University; 2 F.M. Kirby Research Center, Kennedy Krieger Institute 3 Department of Neuroscience, Johns Hopkins University According to the biased competition model of visual attention, when multiple stimuli fall within the visual field: 1) their neural responses compete with each other, and that competition is biased towards the attended stimuli over the unattended (1) ; 2) prefrontal cortex (PFC) has been shown to provide the top-down control to bias processing in the sensory areas (2) . Several studies have shown that PFC can represent abstract, non-sensory information, such as abstract task rules (3) . However, the neural mechanism involved in resolving competition between between task rules is still unknown. 0.90 0.91 0.92 0.93 0.94 0.95 0.96 Congruent Incongruent Accuracy (%) 1.06 1.08 1.1 1.12 1.14 1.16 Congruent Incongruent RT (seconds) Task. 320 total trials (Congruent cue/Incongruent x Semantic task/Phonological task x Color/Shape). Scanning parameters. 3T Philips Intera scanner, T2*-weighted EPI (35 axial slices, TR/TE= 2000/3.7 ms, flip angle = 70°, 2.5 x 2.5 x 2.5 mm 3 voxel size). Participants. Sixteen (12 females) right-handed, healthy adults between 18 and 35 years old, participated the study. All were native English speakers. Error bars represent within-subject variability (95% confidence level) 0.596 0.564 0.50 0.52 0.54 0.56 0.58 0.60 0.62 right IFG right IPS Classifica:on Accuracy n.s. * p<0.05 (group level) 0.552 0.506 0.50 0.52 0.54 0.56 0.58 0.60 right IFG right IPS Classifica:on Accuracy n.s. n.s. Classifying Task Rules (semantic vs. phonological) 1. Training sessions: learn arbitrary pairing between color/shape and task rule 2. Test session (inside of a MRI scanner) 3. Functional localizer Attend Shape + Tiger Shape Block Attend Color + Tiger Color Block Green semantic rule Shape Color Rest Shape Color Rest . . . Rest Rest 8 trials per block 8 trials per block Classifying Context (Attend Color vs. Attend Shape) Triangle phonological rule 0 2 4 6 8 10 Incongruent Cue Congruent Cue Beta values left BA6 (Phonological > Semantic) Semantic Task Phonological Task -1 0 1 2 3 4 5 6 Incongruent Cue Congruent Cue Beta values Left BA47 (Semantic > Phonological) Semantic Task Phonological Task The congruency of the cue did not appear to change the amount of activation in either the relevant or irrelevant task areas (interaction p > 0.05). However, the congruency effect should be further examined in light of the strength of top-down modulation from right IFG. Participants were slower and less accurate in the incongruent trials. Other analyses (not shown) have ruled out the possibility that this effect was modulated by task, attended cue dimension, and response congruency. -4 -2 0 2 4 -3 -2 -1 0 1 2 3 4 5 Enhancement in left BA6 right IFG activity (incon – con) During Phonological Task -6 -4 -2 0 2 4 6 8 -3 -2 -1 0 1 2 3 4 5 6 Suppression in left BA6 right IFG activity (incon - con) During Semantic Task -6 -4 -2 0 2 4 -3 -2 -1 0 1 2 3 4 5 Suppression in left BA47 right IFG activity (incon - con) During Phonological Task -6 -4 -2 0 2 4 6 -3 -2 -1 0 1 2 3 4 5 6 Enhancement in left BA47 right IFG activity (incon – con) During Semantic Task r = 0.79 p<.001 0 2 4 6 8 Congruent Incongruent Activity Task-Relevant Area r = 0.546 p=0.029 d’ = enhancement index (incongruent - congruent) 0 2 4 6 8 Congruent Incongruent Activity Task-Irrelevant Area d’ = suppression index (congruent - incongruent) r = -0.399 p=0.126 r = -0.333 p=0.208 Classifier: Linear support vector machine Leave-one-run-out cross validation results show significantly above chance classifying accuracy for task rule in right IFG, but not for context. Task-relevant area Task-irrelevant area Left BA6 (Phonological Area) Left BA47 (Semantic Area) To test whether the principle of biased competition applies to the selection of competing abstract information, such as selecting relevant over irrelevant task rules. Prediction 1: the source of top-down modulation that resolves competition between rule representations is derived from structures within PFC Prediction 2: Increases in top-down modulation are associated with resolution of competition between task-specific areas. Syllable Concrete or Mayhem Cue: 2s 6 words in total Worship 6/6 Correct Green :: Semantic rule Concrete word (left key) Abstract word (right key) Training Session 1 color-rule pairing (320 trials, ~30 mins) Blue :: Phonological rule 2-syllable word (left key) non-2-syllab. word (right key) Circle :: Semantic rule Concrete word (left key) Abstract word (right key) Triangle :: Phonological rule 2-syllable word (left key) non-2-syllab. word (right key) Training Session 2 shape-rule pairing (320 trials, ~30 mins) Congruency of the cue is manipulated through training. Congruent cues: Incongruent cues: Regions of interest (ROIs) were identified for each participant using functional localizer. Left BA6 and left BA47 were chosen for further ROI analysis based on a previous study (4) . Phonological > Semantic Semantic > Phonological Biased competition predicts the source of top-down modulation that resolves competition between rule representations is derived from PFC. Within the PFC, a contrast of incongruent cue greater than congruent cue showed robust activity in right inferior frontal gyrus (IFG). This area was used as a functionally-defined ROI to explore the interaction between top- down modulation and the activity in task-selective areas. Left BA6 Left BA47 ** * * p<0.01 ** p<0.005 * p<0.05 ** p<0.01 *** p<0.005 ** * *** *** P<0.05 (corrected, cluster level) R L IFG midbrain insula inferior parietal lobule precuneus putamen Incongruent Response vs. Congruent Response BoAle Jus:ce Abstract Non-2-syllab. Concrete 2-syllable Abstract Non-2-syllab. Concrete 2-syllable P<0.05 (corrected) R L Incongruent Cue vs. Congruent Cue postcentral gyrus Poster Session:573.14 Poster #: JJJ13 Tues. AM, Halls B-H References 1. Desimone, R. and J. Duncan (1995). "Neural mechanisms of selective visual attention." Annu Rev Neurosci 18: 193-222. 2.Fuster, M., R. H. Bauer, et al. (1985). "Functional interactions between inferotemporal and prefrontal cortex in a cognitive task." Brain Res 330(2): 299-307. 3.Wallis, J. D., K. C. Anderson, et al. (2001). "Single neurons in prefrontal cortex encode abstract rules." Nature 411(6840): 953-6. 4. Sakai, K. and R. E. Passingham (2006). "Prefrontal set activity predicts rule-specific neural processing during subsequent cognitive performance." J Neurosci 26(4): 1211-8. Cue:0.3s + Delay:0.3s Bottle Target:2s Correct Fdbk:0.5s Trial structure during training: + Fix:1s

Upload: others

Post on 11-Oct-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Testing the Biased Competition Model of Attention in the ...representing abstract information. • The results suggest that “biased competition” serves as a mechanism to guide

P<0.05 (corrected, cluster level)

R L

IFG

insula

SMG

IPS

putamen

R L

•  When there was competition between abstract rules, as in the incongruent trials, participants were slower and less accurate, and this effect was accompanied by increased activity in areas associated with cognitive control.

•  Right IFG serves as the source of top-down bias in response to competition between task rules by enhancing activity in brain areas representing relevant information, rather than by suppressing activity in brain areas representing irrelevant information.

•  The neural pattern of right IFG contains rule information, which provides corroborating evidence that right IFG mediates competition between brain areas representing abstract information.

•  The results suggest that “biased competition” serves as a mechanism to guide the selection of task-relevant abstract information.

Testing the Biased Competition Model of Attention in the Selection of Abstract Task Rules Yi-Shin Sheu1, Susan Courtney1,2,3

1Department of Psychological & Brain Sciences, Johns Hopkins University; 2F.M. Kirby Research Center, Kennedy Krieger Institute 3Department of Neuroscience, Johns Hopkins University

•  According to the biased competition model of visual attention, when multiple stimuli fall within the visual field: 1)  their neural responses compete with each other, and that competition is

biased towards the attended stimuli over the unattended(1); 2) prefrontal cortex (PFC) has been shown to provide the top-down control

to bias processing in the sensory areas(2). •  Several studies have shown that PFC can represent abstract, non-sensory

information, such as abstract task rules(3). However, the neural mechanism involved in resolving competition between between task rules is still unknown. 0.90

0.91 0.92 0.93 0.94 0.95 0.96

Congruent Incongruent

Acc

urac

y (%

)

1.06 1.08

1.1 1.12 1.14 1.16

Congruent Incongruent

RT

(sec

onds

)

Task. 320 total trials (Congruent cue/Incongruent x Semantic task/Phonological task x Color/Shape). Scanning parameters. 3T Philips Intera scanner, T2*-weighted EPI (35 axial slices, TR/TE= 2000/3.7 ms, flip angle = 70°, 2.5 x 2.5 x 2.5 mm3 voxel size).

Participants. Sixteen (12 females) right-handed, healthy adults between 18 and 35 years old, participated the study. All were native English speakers.

Error bars represent within-subject variability (95% confidence level)

0.596  

0.564  

0.50  0.52  0.54  0.56  0.58  0.60  0.62  

right  IFG   right  IPS  

Classifi

ca:o

n  Accuracy  

n.s.  

* p<0.05 (group level)

0.552  

0.506  0.50  

0.52  

0.54  

0.56  

0.58  

0.60  

right  IFG   right  IPS  

Classifi

ca:o

n  Accuracy  

n.s.  

n.s.  

Classifying Task Rules (semantic vs. phonological)

1. Training sessions: learn arbitrary pairing between color/shape and task rule

2. Test session (inside of a MRI scanner)

3. Functional localizer

Attend Shape

+ Tiger

Shape Block

Attend Color

+ Tiger

Color Block Green à semantic rule

Shape Color Rest Shape Color Rest . . . Rest Rest

8 trials per block 8 trials per block

Classifying Context (Attend Color vs. Attend Shape)

Triangle à phonological rule

0

2

4

6

8

10

Incongruent Cue

Congruent Cue

Bet

a va

lues

left BA6 (Phonological > Semantic)

Semantic Task

Phonological Task

-1 0 1 2 3 4 5 6

Incongruent Cue

Congruent Cue

Bet

a va

lues

Left BA47 (Semantic > Phonological)

Semantic Task

Phonological Task

The congruency of the cue did not appear to change the amount of activation in either the relevant or irrelevant task areas (interaction p > 0.05). However, the congruency effect should be further examined in light of the strength of top-down modulation from right IFG.

Participants were slower and less accurate in the incongruent trials. Other analyses (not shown) have ruled out the possibility that this effect was modulated by task, attended cue dimension, and response congruency.

-4

-2

0

2

4

-3 -2 -1 0 1 2 3 4 5

Enh

ance

men

t in

left

BA

6

right IFG activity (incon – con)

During Phonological Task

-6 -4 -2 0 2 4 6 8

-3 -2 -1 0 1 2 3 4 5 6

Sup

pres

sion

in

left

BA

6

right IFG activity (incon - con)

During Semantic Task

-6

-4

-2

0

2

4

-3 -2 -1 0 1 2 3 4 5

Sup

pres

sion

in

left

BA

47

right IFG activity (incon - con)

During Phonological Task

-6 -4 -2 0 2 4 6

-3 -2 -1 0 1 2 3 4 5 6

Enh

ance

men

t in

left

BA

47

right IFG activity (incon – con)

During Semantic Task

r = 0.79 p<.001

0 2 4 6 8

Congruent Incongruent

Act

ivity

Task-Relevant Area

r = 0.546 p=0.029

d’ = enhancement index (incongruent - congruent)

0 2 4 6 8

Congruent Incongruent

Act

ivity

Task-Irrelevant Area d’ = suppression index (congruent - incongruent)

r = -0.399 p=0.126

r = -0.333 p=0.208

•  Classifier: Linear support vector machine

•  Leave-one-run-out cross validation results show significantly above chance classifying accuracy for task rule in right IFG, but not for context.

Task-relevant area Task-irrelevant area

Left BA6 (Phonological

Area)

Left BA47 (Semantic

Area)

•  To test whether the principle of biased competition applies to the selection of competing abstract information, such as selecting relevant over irrelevant task rules.

•  Prediction 1: the source of top-down modulation that resolves competition between rule representations is derived from structures within PFC

•  Prediction 2: Increases in top-down modulation are associated with resolution of competition between task-specific areas.

Syllable Concrete or

Bottle Bottle Mayhem

Growth Justice

Cue: 2s

6 words in total

Worship 6/6 Correct

Green :: Semantic rule Concrete word (left key) Abstract word (right key)

Training Session 1 color-rule pairing

(320 trials, ~30 mins)

Blue :: Phonological rule 2-syllable word (left key)

non-2-syllab. word (right key)

Circle :: Semantic rule Concrete word (left key) Abstract word (right key)

Triangle :: Phonological rule

2-syllable word (left key) non-2-syllab. word (right key)

Training Session 2 shape-rule pairing

(320 trials, ~30 mins)

•  Congruency of the cue is manipulated through training.

Congruent cues:

Incongruent cues:

Regions of interest (ROIs) were identified for each participant using functional localizer. Left BA6 and left BA47 were chosen for further ROI analysis based on a previous study(4).

Phonological > Semantic Semantic > Phonological

Biased competition predicts the source of top-down modulation that resolves competition between rule representations is derived from PFC. Within the PFC, a contrast of incongruent cue greater than congruent cue showed robust activity in right inferior frontal gyrus (IFG). This area was used as a functionally-defined ROI to explore the interaction between top-down modulation and the activity in task-selective areas.

Left BA6 Left BA47

** *

* p<0.01

** p<0.005

* p<0.05

** p<0.01

*** p<0.005

** * *** ***

P<0.05 (corrected, cluster level)

R L

IFG

midbrain insula

inferior parietal lobule precuneus

putamen Incongruent Response vs.

Congruent Response

BoAle  

Jus:ce  Abstract

Non-2-syllab. Concrete 2-syllable

Abstract Non-2-syllab.

Concrete 2-syllable

P<0.05 (corrected)

R L

Incongruent Cue vs. Congruent Cue

postcentral gyrus

Poster Session:573.14 Poster #: JJJ13 Tues. AM, Halls B-H

References 1. Desimone, R. and J. Duncan (1995). "Neural mechanisms of selective visual attention." Annu Rev Neurosci 18: 193-222. 2. Fuster, M., R. H. Bauer, et al. (1985). "Functional interactions between inferotemporal and prefrontal cortex in a cognitive task."

Brain Res 330(2): 299-307. 3. Wallis, J. D., K. C. Anderson, et al. (2001). "Single neurons in prefrontal cortex encode abstract rules." Nature 411(6840): 953-6. 4. Sakai, K. and R. E. Passingham (2006). "Prefrontal set activity predicts rule-specific neural processing during subsequent

cognitive performance." J Neurosci 26(4): 1211-8.

Cue:0.3s

+

Delay:0.3s

Bottle

Target:2s

Correct

Fdbk:0.5s Trial structure during training: +

Fix:1s