Brain activity during transitive and social action observation in adults and

adolescents

M. Lesourd1,2, A. Afyouni1,2, F. Geringswald1,2, L. Raoul1,2, F.Cignetti3, J.

Sein4, B. Nazarian4, J.-L. Anton4, & M.-H. Grosbras1,2

1 Laboratoire de Neurosciences Cognitives (UMR 7291), Marseille, France 2 Fédération 3C, Marseille, France 3 Univ. Grenoble Alpes, TIMC-IMAG, Grenoble, France 4 INT Institut des Neurosciences de la Timone, UMR 7289, Centre IRM Fonctionnelle Cérébrale, Marseille, France

Background and rationale

� Action Observation Network (AON) = set of bilateral brain areas consistentlyengaged during observation of others’ action

Grosbras et al., Human Brain Mapping 2012

Intraparietal sulcus (IPS)

Lateral Occipitotemporal

Cortex (LOTC)

Posterior superior

temporal sulcus (pSTS)

Dorsal/Ventral Premotor

cortex (PMd/PMv)

Inferior Frontal

Gyrus (IFG)

Modulation of AON by the content of action

Iacoboni et al., 2004 Saggar et al., 2014 Centelles et al., 2011

Wurm et al., 2017

MVPA

Decoding social vs

non social action

Social vs individual interactions Social vs Non-social gestures Interacting vs isolated mvts

Development of the AON

� Core nodes of AON are present in children and adolescents (Ohnishi 2004)� Strength and extent of activity increase with age (Shaw 2011, 2012; Biagi 2016)

� From bilateral to more lateralized network (Biagi 2015)

� Occipitotemporal representation become more specific (Scherf 2007; Ross 2014; Peelen 2009; Pelphrey 2009)

� Structural changes in AON regions throughout adolescence (Gotgay 2010, Mills 2012)

� Improvement of social skills (Burnett 2011, Dumontheil 2012) � Inferring intentions of others

� adapting actions to social context, recognizing emotions…

Questions

� Does the representation of action categories in the AON change duringadolescence ?

� Differential activation for different categories of action (object-related; socially-

oriented)

� Fine-grain representation of action categories (coding of social or transitive dimensions)

� More drastic developmental changes in the representation of the « social » dimension

Material and Method

ParticipantsStimuli

Sociality

Transitivity

Tra

nsi

tive

Intr

an

siti

ve

Social Non-Social

Procedure

ST NT

SI NI

24 Adults aged 24-33 (26.6 ± 2.1), 13f

27 Adolescents aged 13-17 (15.0 ± 1.2), 13f

1 bloc = 3 videos (9 sec)

1 run = 20 blocs (6 min)

8 runs

Data analysis I: Univariate analysis

Preprocessing of fMRI data

� Realignement (6-head motion correction)

� Co-registration

� Normalization to MNI space

� Smoothing (5 FWHM)

� Physiological noise (WM & CSF)

� Removal of non-corrected volumes

exceeding FD = 2mm

Data analysis I: Univariate analysis

GLM

� Realignement (6-head motion correction)

� Co-registration

� Normalization to MNI space

� Smoothing (5 FWHM)

� Physiological noise (WM & CSF)

� Removal of non-corrected volumes

exceeding FD = 2mm4 regressors (NI, NT, SI and ST)

Design Matrix (e.g., 1 run)

Preprocessing of fMRI data

Univariate fMRI results

Data analysis II: Multi-voxel Pattern analysis (MVPA)

Preprocessing of fMRI data

� Realignement (6-head motion correction)

� Co-registration

� Normalization to MNI space

� Smoothing (5 FWHM)

� Physiological noise (WM & CSF)

� Removal of non-corrected volumes exceeding FD = 2mm

Data analysis II: Multi-voxel Pattern analysis (MVPA)

GLM

univariate

4 regressors (NI, NT, SI and ST)

Preprocessing of fMRI data

� Realignement (6-head motion correction)

� Co-registration

� Normalization to MNI space

� Smoothing (5 FWHM)

� Physiological noise (WM & CSF)

� Removal of non-corrected volumes exceeding FD = 2mm

Data analysis II: Multi-voxel Pattern analysis (MVPA)

� Realignement (6-head motion correction)

� Co-registration

� Normalization to MNI space

� Smoothing (5 FWHM)

� Physiological noise (WM & CSF)

� Removal of non-corrected volumes exceeding FD = 2mm

GLM

univariate

1 regressor for each trial (4 NI, 4 NT, 4 SI and 4 ST)

multivariate

Preprocessing of fMRI data

Data analysis II: Multi-voxel Pattern analysis (MVPA)

(1) ROI-based: Regions of the AON (Grosbras et al., 2012)

(2) Whole-brain: Searchlight = 12mm radius sphere (Kriegskorte 2006)

• Support Vector Machine (SVM) classifier

IPS/SPL

Left: 213 ± 38 voxelsRight: 172 ± 28 voxels

PMv

Left: 140 ± 23 voxelsRight: 143 ± 25 voxels

pSTS

Left: 72 ± 13 voxelsRight: 75 ± 18 voxels

LOTC

Left: 236 ± 34 voxelsRight: 317 ± 52 voxels

Feature selection (F-test)

IPS/SPL: 104 voxelsPMv: 96 voxels

pSTS: 47 voxels

LOTC: 174 voxels

Data analysis II: Multi-voxel Pattern analysis (MVPA)

• Cross-validation scheme with leave-One-Run-Out

16 betas

16 betas

112 betas

112 betas

8 iterations

8 iterations

Data analysis II: Multi-voxel Pattern analysis (MVPA)

• Cross-validation scheme with leave-One-Run-Out

56 betas

56 betas 8 betas

8 betas

Decoding Social vs Non-Social actions

*** p < .001 (FDR-corrected)

MVPA: ROI-based decoding

Chance level

Adults > Ado

(p = .0098)

Adults > Ado

(p = .006)

Adults > Ado

(p = .0039)

LOTC

left rightPMv pSTS IPS/SPL

left right left right left right

Adolescents

Adults

Decoding Social vs Non-Social actions

MVPA: ROI-based decoding

LOTC

Left Right

PMv

Left Right

pSTS

Left Right

IPS/SPL

Left Right

WithinAdults > Ado

(p = .07)

Adults > Ado

(p = .068)

Adults > Ado

(p = .018)

* p < .05, *** p < .001 (FDR-corrected)Chance level

Decoding Social dimension across transitivity

MVPA: ROI-based decoding

Adults > Ado

(p = .06)

Adults > Ado

(p = .058)

Across

LOTCleft right

PMv pSTS IPS/SPL

left right left right left rightLOTC

left rightPMv pSTS IPS/SPL

left right left right left right

Adolescents

Adults

MVPA: ROI-based decodingDecoding Transitive vs Intransitive actions

Adults > Ado

(p = .015)

LOTC

left rightPMv pSTS IPS/SPL

left right left right left right

*** p < .001 (FDR-corrected)

Chance level

Adolescents

Adults

Decoding Transitive vs Intransitive actions

MVPA: ROI-based decoding

LOTC

Left Right

PMv

Left Right

pSTS

Left Right

IPS/SPL

Left Right

* p < .05, ** p < .01, *** p < .001 (FDR-corrected)Chance level

Decoding Transitive dimension across Sociality

MVPA: ROI-based decoding

Adults > Ado

(p = .001)

Within

Adults > Ado

(p = .03)

Across

LOTC

left rightPMv pSTS IPS/SPL

left right left right left right

LOTC

left right

PMv pSTS IPS/SPL

left right left right left right

Adolescents

Adults

MVPA: Searchlight (whole-brain)

MVPA: Searchlight (whole-brain)

MVPA: Searchlight (whole-brain)

MVPA: Searchlight (whole-brain)

Left IPS/SPL

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

NI

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

NT

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

SI

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

ST

Non-Social Intransitive (NI) Non-Social Transitive (NT)

Social Intransitive (SI) Social Transitive (ST)

*

r = .47 r = .61

MVPA: Multiclass decoding and age effect

Adolescents Adults

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

ST

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

NT

0

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

SI

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

NI

Right IPS/SPL

Non-Social Intransitive (NI) Non-Social Transitive (NT)

Social Intransitive (SI) Social Transitive (ST)

*

r = .39

MVPA: Multiclass decoding and age effect

Adolescents Adults

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

ST

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

NT

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

SI

10

20

30

40

50

60

70

80

90

125 175 225 275 325 375 425

NI

Right pSTS

Non-Social Intransitive (NI) Non-Social Transitive (NT)

Social Intransitive (SI) Social Transitive (ST)

* *

*

MVPA: Multiclass decoding and age effect

Adolescents Adults

Adolescents > Adults

Voxel-Based Morphometry analysis

Adolescents

Adults

Discussion

�AON is already present in adolescents� Sensitivity to the nature of action (sociality and transitivity)

� Fine-grain representations of action category

Discussion

�AON is already present in adolescents� Sensitivity to the nature of action (sociality and transitivity)

� Fine-grain representations of action category

�Adolescents recruit more regions of social brain (mPFC and TPj)

Discussion

�AON is already present in adolescents� Sensitivity to the nature of action (sociality and transitivity)

� Fine-grain representations of action category

�Adolescents recruit more regions of social brain (mPFC and TPj)

�Differences within the AON� Less reliable representation of the content of the action (pSTS and IPS)

Discussion

�AON is already present in adolescents� Sensitivity to the nature of action (sociality and transitivity)

� Fine-grain representations of action category

�Adolescents recruit more regions of social brain (mPFC and TPj)

�Differences within the AON� Less reliable representation of the content of the action (pSTS and IPS)

� IPS/SPL� Age effect for social intransitive actions

� Complex goal oriented behavior and social interaction (Tunik 2007)

� « Beyond grasping »

� Structural difference not associated with age effect

« Social aspect of action representation is still

maturing in adolescents, in bilateral IPS, a core node

of the AON »

Conclusion

Thanks for your attention!