Dissociation of neural circuitry for conscious and

unconscious processing of personally-significant faces.

María A. Bobes

Center for Neuroscience, Cuba

Many different types of information derived from faces:race, sex, emotional-expressions, age, identity

Bruce and Young, 1986Bruce and Young, 1986

Assoc. prospag..

Faces tell us who is who (e.g. good and bad or friends and enemies)

Face identity hasemotional content !

Emotion-from-identity

Face identity also offers socio-emotional information

Tranel, Damasio & Damasio, JOCN, 1995, 7:432

Double dissociation between identity recognition and emotional processing of face identity

Prosopagnosia

Medial orbitofrontal damage

Capgras

Modified Bruce & Young model to accommodate double dissociations

(Ellis et al)

Assoc. prospag..

Me

dia

l O

rbit

ofr

on

tal

Limitations of purely cognitive models

• Theories are under-constrained• They have poor contact with neurophysiological data• The posited pathways are not informed by

neuroanatomy• Early attempts to make the connections were heavily

criticized because they were based on incomplete data & were difficult to test (recommendation then to ignore anatomy)

FFAOFA

Hasson, Harel, Levy, & Malach, Neuron, 2003

= >

STS

Posterior Superior Temporal Sulcus(pSTS)

Core Area of the Face Processing System

(Haxby et el, 2000) Extended system for familiar faces?

What is unclear• Are there additional areas involved?

• What are the connections of these areas

• What is the functional contribution of each area

• How brain damage disrupt this circuitry

fMRI correlation

DTI. Tractography

Lesions causal inferences

fMRI Event related design

St Duration: 1s

SOA: variable between 4 and 6 s

3 sessions. Each session:•15 acquaintance faces,•15 learned faces, 30 unknown faces•45 houses.

Target (40 %)Standard (60 %)

1.5 Tesla

TR= 2

16 slices

Statistical Analysis: SPM5+ in-house scripts

fMRI analysis

• ROI localization of areas responding to faces

(faces > houses)

• Finite impulse response estimations for each stimuli condition in each ROI

• Permutation t test over the FIR. (Time characterization)

•FFA (L/R)

•OFA (L/R)

•AIT (L/R)

•Post. Cing.(L/R)

•Medial Orb.fr. (L/R)

•Insula

•Dorsolateral Frontal (L)

•Inferoparietal (L)

Activation for faces >houses

Regions of Interest ROIStructural face processing

Emotional processing

Attention

working memory

Semantic processing

FFA d

OFA d

FFA d

OFA dOFA dAcq vs 0

Unfam vs 0

Acq vs Unfam

OFA d

FFA d

OFA d

Fam vs 0

Desc vs 0

Fam vs Desc

FFA d

OFM d

Unknown faces > Houses

FFA

mOFC

OFA

STS

acquaintance > unknown faces

PC

Group analysis

AIT

To Summarize

• BOLD responses to faces of acquaintances were of larger amplitude and longer duration than to faces of strangers in all ROI, especially in the right hemisphere.

• This differential responsiveness increases as one moves from extrastriate areas to PC to lateral and medial prefrontal cortex.

• There are striking differences in BOLD time-courses in different ROI and interestingly for different conditions in the same ROI (excludes local vascular factors or regional differences in the coupling of neural activity) .

• This suggests that different BOLD time-courses could reflect distinct neural patterns of activity.

Basser et al., 1999Mori et al., 1999

Attempt to ‘connect’ voxels

on basis of directional

similarity of coincident

eigenvectors

Catani, Brain, 2005

Classical DTI tractography: tracts defined by ROIs in white matter:-Good “in vivo” reproduction of tracts found by “post-mortem” dissection -Unfortunately relationship with functional areas (e.g. OFA, FFA) not precise-Gets “lost” or stops in GM

SLF

Probabilistic DTI-tractography seeded with fMRI data

• A Monte Carlo simulation (n=200) is run at each seed voxel and at every step of the path construction a direction randomly drawn from a distribution).

• Validity measure calculated for all paths.• Seed ROIs can be placed in GM• Paths can penetrate GM• Here we used clusters of fMRI activation specific to

faces as seeds for the tractography• Average of validity measures for all paths coming

from the same ROI and reaching any voxel was used as an index of connectivity between the ROI and the voxel.

Seed ROI defined by fMRI

OFA

FFA

Voxelwise, T-test, n=30, of connectivity index, FDR=0.05

Additive color map

(pSTS)mOrF

Modified model of the Face Processing System

SLF?

CP

lOrF?

IFL

IFOF

Summary of new findings from normal subjects

1. Core areas have larger BOLD responses to familiar faces but ERPs elicited in these areas are not affected by familiarity: Suggests feedback from later activated areas.

2. The extended face processing system for familiar faces includes the posterior cingulate and medial orbitofrontal cortex.

3. DTI-tractography seeded by fMRI data suggest a massively parallel face processing system, less sequential than posited in original models.

4. V4, OFA and FFA all feed in parallel into the Inferior longitudinal fasciculus and the Inferior Fronto-occipital fasciculus.

• Are all these areas necessary ?

• Need for lesion studies

• Brain injury / bilateral subdural hematomas (trauma).

• Neuropsychology: normal except face recognition• Basic face processing relatively intact (Benton, Bruyer and Warrington test batteries).•Dense associative prosopagnosia (zero recognition of famous people or acquaintances), cannot learn new faces

Severe bilateral damage to ventral occipito-temporal cortex

Case F.E. . (69 year old male)

Person identification relies on voice, gait, clothing

LR

nsp<0.028

8/14 (57 %)

12/14

(86 %)

10/14 (71 %)

Which face is more trustworthy ?

Which face is nicer ?

Which face is more trustworthy ?

Which face is nicer ?

Forced choice face selectionForced choice face selectionForced choice face selectionForced choice face selection

Bobes et al, COGNITIVE NEUROPSYCHOLOGY, 2004, 21

Unknown Acquaintance (family)

p<0.008

Is more familiar ?

Implicit Implicit Overt

Skin Conductance Response (SCR)

F.E. presented larger skin conductance responses to familiar than to unfamiliar faces.

Bobes et al, COGNITIVE NEUROPSYCHOLOGY, 2004

Acquaintances unfamiliar

Voxel-wise Hotelling T2 test, using vector of intravoxel GM, WM and SCF concentration, contrasting FE vs. 237 control

FDR threshold q<0.01

RL

% Damaged voxels in cortical regions defined by AAL atlas

Patient FE. ER-fMRI Acquaintances, Unfamiliar faces, Houses

• ROI (faces > houses)

• Finite impulse response estimations for each stimuli condition in each ROI

• Statistical comparison of the patient´s response to that of the sample of normal controls (Crawford et al, 2006)

OFA OFAFFA

FFA

OFA

OFA

Left Right

x=44 y=-78z=-14

1

PC CA mOFC

mOFCmOFC

x=0 y=34z=10

Left Right

FE:Left mOFC

Controls> FE

Voxel-based tractography (OFA seeds) deficit analysis: voxel-wise ttest, FE vs 30 controls, FDR q=0.01

Deficits in the territory of the right and left ILF

Method limited by large variance of values in some voxels control subjects

>70 % controls

Only F.E.

Overlap F.E. & controls

Combined DTI/fMRI: Probabilistic Tracts with group OFA seeds for controls and individual OFA seeds for F.E.

IFOF Apparently preserved (more on the on the left )Lower branches of IFL leading to AIT cortex and MTL are more affected

(pSTS)mOrF

SLF?

CP

lOrF?

IFL

IFOF

Residual face processing system in FE

X

X

X

X X

(pSTS)mOrF

SLF?

CP

lOrF?

IFOF

Residual face processing system in FE

XX X

(pSTS)mOrF

ERPs ?

Time course of activation

SLF?

CP

lOrF?

IFL

IFOF

Newly-Learned FacesFaces of unknown persons artificially learned in the lab

Unique Identity

Memory engrams

Unique identity based on face structural information

Affective significance

Semantic Information

Similar visually familiarity

Different emotional value

Acquaintance faces Faces of relatives and very close friends

ERP paradigmERP paradigm

120 channels

0.05-30 Hz

10000Onset latency

St Duration: 1sSt Duration: 1s

SOA: 2 sSOA: 2 s

Two blocks : Random presentation of 16 target known faces and 75 standard unknown faces .

In one block known faces were acquaintances and in the other, newly-learned faces.

Task: To discriminate between familiar (acquaintance or learned) and unfamiliar faces by pressing a key

Subjects: 15 normal adults (7 F and 8 M), ages between 21 and 32

P300P300N170 N170

An iAn index of early ndex of early structural analysis of structural analysis of faces faces

An index ofAn index of the availability of relevant the availability of relevant memories for people.memories for people.

P3b. (parieto-central) •P3a . (frontally distributed)

Learned Faces (LF) in normal subjects.

Target LFStandard UF

•No difference in N170 (structural processing)

•P3b. Conscious processing of familiarity

Acquaintance faces (AF) in normal subjects

Target AFStandard UF

•No difference in N170 (structural processing)

•P3b. Conscious processing of familiarity

•Additional P3a. Emotional processing of identity

A

Newly-learned

500-580ms

B Acquaitances460-570 ms

CAcquaitances300-380 ms

•Temp-parietal• Polo temp• Frontal inf.• Frontal dorsal

• Temp-parietal• Polo temp• Frontal inf.• Frontal dorsal

• OFM• Amigdalas• Polo temporal, • Insula

A) P3b. Newly-learned

B) P3b. Acquaitances

C) P3a. Acquaitances

Current source stimation BMA mode.l Average Brain MNI .

Summarizing• The effect of familiarity appears after 300 ms

once the structural processing (indexed by N170 has been completed)

• Learned faces evoke only the parietal positivity

• Acquaintance faces evoke two different components: a large positivity widely distributed and maximal at parietal sites, and an earlier positivity restricted to frontal sites, which is probably generated at OFM and amygdalas.

ERPs. P300

0 % hits

Bobes et al, COGNITIVE NEUROPSYCHOLOGY, 2004, 21

• P3a adicional: Procesamiento emocional de la identidad• P3a adicional: Procesamiento emocional de la identidad

Target AFStandard UF

•No difference in N170 (structural processing)

•An early and fronto-central P3 (P3a). Emotional processing of identity

•No P3b. (Conscious processing of familiarity)

•Metodo. BMA

•Volumen. RMN sujeto FE

OFM, amigdalas, polo temporal

FE. Estimación de fuentes intracraniales. P3a o P3b?

Las fuentes generadoras de P300 en FE coinciden con las de P3a en normales.

ERP• The processing of emotion from identity does

not necessarily occur before identity have been consciously detected.

• Covert recognition of familiar faces in prosopagnosia evoked the P3a component generated in emotional system

Conclusions1. FFA is not necessary for covert recognition of

familiar faces which could be enabled by OFA. Previous work (Rossion et al) has show that FFA activation is not sufficient for familiar face recognition.

2. OFA and med OrbitoFrontal cortex serve as hubs for covert (unconscious) recognition

3. Prosopagnosia not only due to lesion of “core face areas” but also a disconnection syndrome (IFL disruption) that blocks conscious recognition

Conclusions4. Unconscious recognition could be mediated by

the Inferior Occipito-Frontal Fasciculus or by more indirect pathways via the SLF or the Cingulate fasciculus.

5. Multimodal neuroimaging of neuropsychological cases is a powerful constraint for models of cognitive processing.

6. More effort is needed to develop of statistical techniques for multi-modal image comparison of singles cases to normal controls.

Research team Habana, Cuba• Maria A. Bobes• Lorna Garcia• Yasser Iturria• Ileana Quiñones• Yusniel Santos• Lester Melie• Pedro Valdes Hernandez • and many others from CNEURO

Medellin, Colombia• Francisco Lopera• Jose Ascencio