2P044 六基本表情の認知能力間の相関構造(2003年度 日本基礎心理学会第22回大会優秀発表賞)

The Japanese Psychonomic Society

NII-Electronic Library Service

The JapanesePsychonomic Society

11ie

Jopanese

Jbumal

o,.f

Ps,yehonomic

Science

2004,

VvL

23,

A'o,

1,

107-108

Summaryof

Awarded

Presentation2P044

The

interrelationship

theof

recognition

of

facial

six

basic

emotions

expression

of

Atsunobu

SuzuKi,Takahiro

HosmNo,i)

The

Uitiversit.v

qf

Tokyoand

Kazuo

SHIGEMASU

*

An

interrelationship

in

the

recognition

of

facial

expressions

of

the

six

basic

and specific

emetions

has

been

inferrcd

from

neurological

cases

who

have

displayed

a

disproportionatc

impairment

in

recognizing

them.

The

_previous

reports,

however,

nre s･,ulnerable

to

thc

criticism

that

the

apparent

emotion-specific

inipairments

are merety artifacts

due

to

the

various

levc]s

of

difficulty

in

rccognising

the

different

emotions,

Wc

propose

a

new

methed

for

the

assessment

of

recognition of

facial

expression.

In

thts

mcthod,

the

dithculty

artifact$

arc

fully

controlled

by

means

of

morphing and

item

response

theory

CIRT),

We

used a morphing

technique

to

create

intermediate

facial

expre$$ions

which

would

have

various

levels

of

diraculty

for

recognising each ernotion.

By

using

IRT,

the

sensitivity

to

the

expressions

was

estimatcd

by

taking

account of

differential

diMculty

leve]s,

The

results revealed

that

a

sen$itivity

to

happiness

alone was not

correlated

with

a

sensitivity

to

the

other emotions, and

indicated

that

the

recognition

ot

happiness

was

independent.

'

i

Key

words:

facial

expression recognition,

basic

ernotions, morphing,

item

response

theory

<IRT}

It

is

well-known

from

the

literature

that

there

is

a

moderately common, cross-cultural ability

to

recog-nise

facial

expressions

which show at

least

a

subset

of

emotions,

the

so-called

basic

emotions:

happiness,

surprise,

fear,

anger,

disgust,

and

sadness.

The

issue

of

an

interrelation$hip

betwcen

the

recognition of

the

six

basic

emotions

is

controversial,

Neuropsy-chological research reports

havc

indicated

that

focal

brain

damage

can causc a

disproportionate

impair-rnent

in

the

recognition of

facial

expressions of spe-ciflc emotions,

especially

fear

and

disgust.

The

result

suggests

that

there

are neural

substrates

that

are

specialized

for

rccognizing certain emotions,

A

broad

deficit

in

recognizing

emotions

whieh ure not

happiness

atso

indicates

a

major

differentiation

be-tween

the

recognition of

happiness

and

the

other

emotions,

Thc

previous

reports,

however,

have

been

criticized

for

a

lack

of control over

the

different

levels

of recognition

difficulty

across

the

emotions.

*

_Department

_of

_Cognitive

_and

_Behavioral

ence,

Graduate

School

of

Arts

and

Sciences,

Thc

University

of

Tok},o,

3-8-1

Komaba,

Meguro-ku,

Tokyo

153-8902

i)

_Takahiro

_Hoshine

_is

_{now at}

_the

_Division

_of

Humanities

and

Social

Sciences,

Departmcnt

of

Interdi$ciplinary

Statistics,

The

Institute

of

Statistical

Mathernatics.

The

critics

have

argued

that

the

apparent

emotion-specific

impairments

are merely artifacts

that

are

due

to

differentia]

difilculty

levels

across

the

cmo-tions,

The

present

article

_proposes

a new method

that

is

designcd

to

assess

the

recognition

of

emotiona]

facial

expression as

"the

sensitivity

to

basic

emotions

indi-cated

by

facial

expressions",

or

simply

"sensitivity

score$".

In

this

method,

_participants

arc asked

to

rate

the

emotional

intensity

ol

morphed

iacial

expres-sions,

and

their

scnsitivity scores are estimatcd

by

app]ying

item

response

theory

CIRT)

to

their

rating

responscs.

Morphing

can

_provide

stimuli with

vari-ous

levels

of recognitien

difliculty

for

each

emotion.

An

elaborate

psychometric

measurement

of

IRT

en-ables

the

estimation

of

sensitivity

scores,

depending

both

on

the

rating

responses

of

the

participants

and

on

the

_properties

of

the

stimulus

which

affcct

the

recognition

difflculty

(Embretson

&

Reise,

2000).

[n

this

study,

by

using

the

new method, we

ex-plored

the

relationship

of

recognition of

the

six

basic

emotions shown

in

facial

expressions

without

the

confounding

factor

of

difficulty

artifacts.

Method

undergradu-The Japanese Psychonomic Society

NII-Electronic Library Service

TheJapanesePsychonomic Society

108

_The

_Japanese

_Journal

of

Psych

ates,

of

whorn

340

were

males,

with an

age

range

of

18

to

31

_years.

Taskandstimuli

The

participants

were

tested

together

in

a

large

classroom

with a

large

screen at

the

front,

Each

of

36

black-and-white

_photographs

of

facial

expressions was

_presented

on

the

screen

for

25

seconds.

During

the

presentation

of each stimulus

the

participants

were askecl

to

iate

the

emotional

inten$ity

of each

facial

expression with respect

to

the

six

basic

emotions.

The

ratings

were

made

on a

6-point

scale,

from

O

{not

at all)

to

5

(very

much),

Six

of

the

36

_photographs

were

_prototypical

im-ages

of

the

six

basic

emotions,

posed

by

a

Japanese

woman,

Each

of

the

other

30

stimuli were morphed

images

of

two

different

_prototypical

_photographs,

For

each of

_the

15

_possiblc

_pairs

of

the

six

proto-types,

two

morphs were created

by

b]cnding

in

the

proportions

of

60:40,

and vice versa.

The

order of

_the

36

stimuli

wa$

quasi-randomized

with

the

constraint

that

any

two

images

containing

the

same

_prototypical

expression were not

_presented

consecutively.

IRT

model

The

scores of

the

sensitivity

to

any

specific

basic

emotion were estimated

by

applying a

graded-response

model

to

the

participants'

rating

re-sponses

to

11

fa

¢

ial

images

containing

the

corre-sponding emotion.

Results

Thc

IRT

estimates of

the

sensitivity scores showed

considerable variability

among

the

participants,

and

their

distributiens

displayed

a unimodal and roughly

symmetrical curve

for

each emotion.

This

result

indicated

that

the

sensitivity scores

were

net

dis-torted

by

diMculty

factors,

The

correlations

of

the

sensitivity scores of each

basic

emotion

are

shown

in

Table

1.

The

sensitivity

scores

of

the

five

basic

emotions,

excluding

happi-ness,

were

significantly

and

_positively

correlated

with each other.

The

sensitivity score

of

happiness,

however,

was

not

cerrelated

with

that

of

any

other

emotlon,

We

conducted an exp16ratory

factor

analysis

to

further

examine

the

correlation structure of

the

six

onomlcScience

VoL

23,

No.

1

Table

1

Correlation

of

the

sensitivity sceres

HA

su

FE

AN

DI

suFEANDISA

O.O09

O.O06

O.O17･-O.O17

O.077

0.366*O.624*

e.351*

O,550*

O.412*

O.511*

O.417*o.sle*O.567*

O,430*

Nbte,

HA;happiness,

SU=surprise,

FE=fear,

AN

=: _anger,

DI=disgust,

SA

= _sadness.

*p<O.OOL

sensitivities,

A

scree

_plot

revealed

that

the

first

factor

alone

had

an

outstanding

eigenvalue

{2.9),

when compared

to

the

following

factors

(O.4-1.0),

indicating

that

a

one-factor

model

was

appropriate.

The

variance

explained

by

the

first

factor

was

40%,

and

the

factor

loadings

were

O.027

for

happiness,

O.770

for

surprise.

O.525

for

fear,

O.773

for

anger,

O.683

for

disgust,

and

O.697

for

sadness.

This

result

rneant an aimost null

factor

loading

for

happiness.

Discussion

Analyses

of

the

correlation structure of

the

sensi-tivity

to

different

ernotions

confirmed

the

indepen-dence

of

a

recognition

sensitivity

for

happiness,

rath-er

than

acomplete

dissociation/association

of

recog-nition

sensitivity

for

each

emotion.

AIthough

previ-ous

studies

which

reported

an

independence

for

the

recognition

of

happiness

have

been

attacked

for

not controlling

diMculty

artifacts,

the

_present

study

ap-pears

to

have

contrelled

these

artifacts

successfully,

because

the

distributions

of

the

sensitivity scores are approximately normaL

Because

happiness

is

the

only completely

positive

emotion of

the

six

basic

emotions, and

the

others are mostly negative,

the

finding

suggests an orthogonality

bctween

_positive

emotion

recognition

and

negative

emotion

recogni-tion,

It

would

be

desirable

to

use several other

positive

emotions

in

any

future

research

which

at-tempted

to

establish

this

orthogonality.

References

Embretson,

S.E.

&

Reise,

S.P.

2000

item

resPonse