Encoding Activity and Face Recognition · TABLE1...
Transcript of Encoding Activity and Face Recognition · TABLE1...
E ncod ing A ctiv ity an d F ace R ecogn ition
C hristia n C o in and G uy T ibe rgh ien
Institu t de Sciences Cogn itives, Lyon, France
A series o f stud ies conducted ove r the past 20 years have explo red the effec ts o f
va r iou s ta sk s on recognition m em ory for faces. M em ory for faces appears be tte r
w hen the s tudy task involv es judgem ents abou t an abs tract tra it ra the r than a
phy sica l feature . T he va riou s si tua tion s in w hich these resu lts w ere ob tain ed raise
im po rtant m e thodo log ica l ques tions rega rding the lea rning cond ition s, w he the r
inc identa l o r in ten tiona l, and th e du ration o f expo su re to the s tim u lus du ring th e
study phase. W e con sid er h ere tw o a ltern ative exp lana tion s for the repo rted results .
O ne conce rns dep th of p ro cessing and the o the r the oppo si tion be tw een com ponen t
and ho list ic pro cess ing . Po ss ib le st ra teg ie s fo r im p roving face recogn ition
pe rfo rm ance a re cons ide red.
INTRODUCTION
W e can read expression from a face presented w ithin 20m s (S im pson &
C randall, 1972) and recogn ise form er schoo lm ates at about a 90% recogn ition
rate 35 years a fter g radua tion (Bahrick , Bahrick , & W ittlinger, 1975). In the past
20 years, num erous stud ies have explo red the influence of encod ing instruc tions
on fac ia l recognition . M ost o f the studies exam ined here , inc lud ing the m ost
recen t (e .g . Terry, 1993), used an experim en ta l parad igm sim ila r to that o f the
p ioneering study reported by B ower and K arlin (1974). S tud ies em ploy ing
adu lts (e.g . B ower & K arlin , 1974 ; C lifford & Prior, 1980 ; M cK elvie , 1991 ;
Patterson & B adde ley , 1977 ; Sporer, 1991 ; W ells & Turtle , 1988 ; W inograd ,
1981) show tha t the requ irem ent fo r persona lity trait judgem ents, such as
likeab ility , y ie ld h igher recogn ition ra tes tha t judgem ent abou t size of a fea ture
or gender. S im ilar patterns of resu lts have been ob tained w ith ch ildren (7 to 14
years o ld : C arey , D iam ond , & W oods, 1980), w ith o lder sub jec ts (50 to 70 years
M E M O R Y , 1 99 7 , 5 (5 ), 5 45 ±56 8
R eq u es ts f o r re pr in ts sh ou ld b e sen t to C h ris tia n C oin , In st itu t de Sc ien ce s C o gn itive s , U nive rs iteÂ
C lau d e B e rn a rd , 8 av en u e R ock e fe lle r, 6 93 73 L y on ceÂdex 08 , F ra n ce .
T h is w ork w as co m plet ed w h ile th e fir st a u tho r w as o n a p os tdo c to ra l fe llow sh ip a t the
M assa chu sett s In s titu te o f T echn o log y , C am b rid g e, M A . P a rts o f th is w ork w e re sup po rte d b y a
Fu lb rig h t g ra n t to the firs t au th o r. W e w o uld l ik e to than k A lan H e in and an an on y m o u s rev iew e r fo r
h elp fu l co m m en ts on an ea rlie r d ra ft o f th is a rti c le , a nd W en dy M cK enz ie and D av id Jam es fo r th e ir
re ad ing of the E ng lish ve rs io n o f th is a rti c le .
Ó 19 9 7 Psy cho lo gy P re ss L td
o ld : W arring ton & A ckroyd, 1975) and w ith am nesic K orsakoff pa tien ts (B iber
e t al., 1981). Th is paper w ill rev iew and eva lua te the various procedures
em ployed in these stud ies.
T here are four categories o f encod ing instructions fo r the study phase of the
experim en ts review ed here :
1 . S tandard instruc tions in w hich sub jects a re to ld on ly that a reten tion test
fo llow s the study phase .
2 . Instruc tions to a ttend to physica l charac teristic s v iew ed d irec tly (e .g .
facia l featu res and accessories like glasses), or inferred (e.g . gender, race, w eight
and he igh t).
3 . Instruc tions to regard globa l aspects o f the face (e .g . expression , shape),
pe rsona lity tra its (e .g . likeab ility and inte lligence). T hese instructions are
som etim es re la ted to identifying activ ities o f the figure (e .g . p rofession, hobby),
o r to relig ious or po litical pre ference .
4 . Instruc tions to choose the m ost d istinctive fac ia l featu re am ong a set of
facia l featu res.
T aken toge ther, the resu lts show that judgem ents o f personality tra its y ie ld
be tte r recogn ition perform ance than those of physical traits. The da ta provided
in Tab le 1 a lso dem onstra te that in ferential judgem ents im prove distingu ish-
ab ility (d ¢ ) am ong them . T he dec ision criterion ( b ), appears unaffec ted by the
encod ing cond itions (see a lso Shapiro & Penrod , 1986) a lthough the da ta
supporting th is conc lusion is sparse . T he faces searched for their m ost d istinctive
physica l fea ture am ong a se t designa ted by the experim en ter were better
rem em bered than those judged w ith respec t to a sing le physica l featu re (C ourtois
& M uel le r, 1 979 ; Sp ore r , 19 91 ; W inograd , 1 981 ). F ina lly , re cogn i tio n
perform ance w as not differen t fo llow ing judgem ents about personality tra its
than after judgem ents abou t the m ost d istinctive fac ial fea ture (D aw & Parkin ,
1981; D effenbacher, Leu , & B row n, 1981; Parkin & G oodw in, 1983 ; Park in &
H ayw ard , 1983 ; W inograd , 1981).
METHODOLOGICAL ASPECTS
B efore a ttem pting an exam ination of the proposed exp lana tions fo r these resu lts
it is im portant to no te the differences in the m ethodo log ies that underlie them . In
som e procedures sub jec ts w ere to ld that a recogn ition test w ould fo llow the
study phase . In o ther experim en ts this info rm ation w as no t g iven . A dditiona lly ,
exposure duration of the target face varies am ong the stud ies. Possib le
consequences of such varia tions m ust be ana lysed . In particu lar, the subjec ts
could adopt a particu lar encod ing strategy w hen they know that a test w ou ld
fo l lo w . A lte rn a tive ly the ir no rm al s tra te gy m ig h t be un a l te red by th e
instruc tions. F ina lly, if d iffe ren t judgem ents have d ifferent p rocessing tim es
546 COIN AND TIBERGHIEN
TABLE 1Face Recognition Performance as a Function of Encoding Instructions
H its F A d ¢ b
A u th o r(s ) F P F P F P F P
B ib e r e t a l. (19 8 1) .6 5 < .8 1 Ð Ð Ð
B lo o m & M ud d (1 9 91 ) .5 6 < .7 6 Ð Ð Ð
B o w er & K arlin (1 97 4 ; E x p . I) .6 0 < .8 1 Ð Ð Ð
B o w er & K arlin (1 97 4 ; E x p I I) .5 6 < .7 6 Ð Ð Ð
C are y e t a l. (1 98 0 )a
.8 0 < .8 8 Ð Ð Ð
C o urto is & M ue lle r ( 19 79 ) .6 9 < .8 4 .2 1 > .1 2 1 .5 < 2 .4 2 .1 ± 2 .7
D evine & M a lpa ss (19 85 ) Ð Ð .8 6 > .8 0b
Ð
G aon ac ’ h & G ib o in (1 97 7 ; E x p . I) .8 5 < .92c
Ð Ð Ð
G aon ac ’ h & G ib o in (1 97 7 ; E x p . II) .7 8 < .87c
Ð Ð Ð
L igh t e t a l . (1 97 9 ; E x p . I)d
.5 4 < .6 5 .2 9 > .2 1 .6 6 < .7 5e
Ð
L igh t e t a l . (1 97 9 ; E x p . II)d
.5 5 < .6 4 .2 8 > .2 1 .6 6 < .7 6e
Ð
M ay es e t al . (1 98 0 )a
.4 2 < .5 0 Ð Ð Ð
M cK e lv ie (1 97 8) .8 3 < .8 8 Ð Ð Ð
M cK e lv ie (1 98 5) .8 2 < .8 7 Ð Ð Ð
M cK e lv ie (1 99 1) .6 6 < .9 0 Ð Ð Ð
M u e ller e t a l. (1 97 8 ; E x p . I)f
.5 8 < .6 6 .2 0 = .1 9 1 .1 < 1 .4 1 .7 = 1 .4
M u e ller e t a l. (1 97 8 ; E x p . II ) .4 8 < .6 7 .2 5 = .2 3 .6 9 < 1 .2 1 .5 = 1 .2
M u e ller e t a l. (1 97 9 ) .6 3 = .6 9 .2 2 > .1 5 1 .2 < 1 .7 1 .8 = 2 .7
M u e ller & W he rry (19 8 0) .5 2 < .7 8 .16 ± .15 1 .2 ± 1 .9 3 .0 ± 1 .5
Pa t te rson & B ad d e ley (1 9 77 ) .7 6 < .8 2g
Ð 1 .5 < 1 .9 Ð
Sm ith & W in og ra d (1 97 8 )h
.7 5 < .8 6 .2 1 = .2 2 1 .5 < 1 .9 1 .0 = 0 .9
Sp o rer (1 9 91 ) .5 6 < .6 6 .30 ± .26 .6 9 < .7 7b
Ð
Strin ge r (1 98 7 ) .5 1 < .6 7 Ð Ð Ð
Strnad & M u e lle r (1 97 7)i
.4 6 < .5 5 .2 6 < .3 4 Ð Ð
T erry (1 99 3 ; E x p . I) .7 4 < .8 1 .1 6 = .1 6 2 .4 = 2 .7 5 .1 = 4 .4
T e rry (1 99 3 ; E x p . II) .6 5 < .7 0 .2 4 > .2 0 1 .4 > 1 .8 3 .0 = 3 .3
W arri ng to n & A ck roy d (1 97 5 ) .80 < .86c
Ð Ð Ð
W ells & H ryc iw (1 98 4 ) .2 5 < .5 5 Ð Ð Ð
W ells & T u rtle (1 9 88 ) .5 4 < .7 8 Ð Ð Ð
W ino g ra d (1 97 6) .66 < .75c
Ð Ð Ð
W ino g ra d (1 98 1; E x p . I) .6 4 < .7 3 .2 5 > .1 6 1 .0 < 1.6e
Ð
W ino g ra d (1 98 1; E x p . II ) .7 0 < .7 6 Ð .8 5 < .8 8e
Ð
F = jud g em ent ab ou t a fa c ial fe a tu re ; P = ju d gem en t abo u t p e rson a lity tra it . D a sh e s in d ic at e n o
d at a av a ilab le . T h e d iffe re n ces a re s ig n if ic an t a t th e .0 5 lev e l a t le a s t; < in fe rio r; > sup e rio r; = n o
s ig n i fic an t d i ffe re nce; ± n o diffe re nce m en tion ed .aD a ta e s tim a ted a s a c cu ra te ly as p o ss ib le fro m a f igu re in th e orig ina l a rti c le .
bT he sens i tiv ity is
exp re ssed b y th e A ¢ es t im a te .cH its p lu s Fa lse P os it iv e s.
dD a ta o b ta in ed fro m typ ica l an d un u su a l
fa c e s.eT h e sens it iv ity is exp re ssed by th e A g e s tim a te .
fD a ta o b ta in ed fro m tw o g ro up s of su b je ct s (5
and 1 5 secon ds ex po sure du ra tion).gS ign ifi c an t a t the .1 0 leve l.
hD a ta ob ta ined fro m tw o g ro up s o f
sub jec ts (18 ±52 y ea rs o ld an d 5 0±8 0 y ea rs o ld ).iD a ta ob ta ined fro m tw o g rou p s of sub jec ts
(in c id en ta l an d in ten t io na l le a rn in g).
547
and exposure dura tion is constan t, the e ffec ts of exposure dura tion m ay be
confounded w ith the type of encoding instruction .
Control Group, Incidental Learning, and IntentionalLearning
Studies designed to exam ine the effects o f encoding instructions on fac ial
rec og n ition va ry in th e ju dg em en ts req u ired . T h is h as tw o im p o rta n t
consequences. F irs t, they relate on ly to a lim ited se t o f strateg ies fo r face
encod ing em ployed outside labora tory settings. Norm ally face encod ing is
conduc ted spon taneously and is free of im posed judgem ents (a t least o f the k ind
d iscussed here). Second , com parison betw een stud ies is difficu lt because of the
lack of a com m on basis, in add ition to unavoidab ly varying procedures be tw een
experim en ts. For exam ple , it is d ifficu lt to draw a para lle l betw een recogn ition
perform ance obta ined after judgem ents abou t sex and about likeab ility (e .g .
B ow er & K arlin , 1974), and those obta ined after judgem ents abou t w eigh t and
in te lligence (e .g . M uelle r, B ailis, & G oldstein , 1979).
In these studies, th ree types of lea rning were used :
1 . S tandard in ten tional lea rn ing in w hich sub jec ts a re instruc ted only to
exam ine the stim uli during the study phase. T hey are w arned abou t the
subsequen t test phase .
2 . N on-standard inten tional learn ing in w hich sub jec ts m ake judgem ents
during the study phase . They are warned abou t the subsequen t tes t phase .
3 . Inciden ta l lea rn ing in w hich subjects m ake judgem ents w ithou t know ing
abou t the subsequen t recogn ition test.
In order to assess the effects of d ifferen t encoding instruc tions w ith regard to
the standard lea rn ing parad igm (basic leve l),1
the lea rn ing fac tor m ust no t be
confounded w ith the type of judgem ent fac tor m ade during the inspec tion phase
(e .g . Dev ine & M alpass, 1985 ; Sporer , 1991). In these experim en ts the learn ing
during the inspection phase is in ten tiona l fo r one group , and inciden ta l w ith
judgem ents to be m ade for the other g roup . Thus, it is no t possib le to com pare
the effec t o f standard in ten tional lea rn ing recogn ition w ith the effec t of
in ten tiona l learn ing w ith judgem ents. H ow ever, B ow er and K arlin (1974)
show ed tha t fac ia l recognition w as superio r fo llow ing judgem ents abou t the
honesty of a face than after judgem ents abou t its gender, both in an inciden ta l
learn ing cond ition (Experim en t 1 ; see also C arey et a l., 1980 ; Pa tte rson &
1T h e ba s ic leve l co n s ide re d h e re is the con d ition w h e re the sub jec t d oe s no t m ake ju d gem en ts
d u rin g th e s tu d y p ha se and is n o tifie d o f th e sub sequ en t te st ph a se . P e rh aps it w ou ld b e p re fe ra b le fo r
the b a s ic lev e l to b e th e con d ition o f in c id en ta l le a rn in g in w hich n o ju dg em ents a re m ade . H ow ev er,
i t m u st be adm itted th a t it i s d if fic u l t to c re a te su ch a co nd it io n .
548 COIN AND TIBERGHIEN
B add e ley , 1977 ; T erry , 1 993) an d in an in ten tio na l lea rn ing con d ition
(Exper im en t 2 ; see Sporer, 1991 ; W ells & T urtle, 1988). The sam e pattern of
resu lts w as ob ta ined by Strnad and M ueller (1977) w here inciden tal and
intentional lea rn ing cond itions w ere com bined w ith in the sam e experim en t.
Judgem ents based on the m ost d istinc tive featu re do no t lead to better
recogn ition perform ance than judgem ents about the infe rred characte r o f the
face in the inciden ta l lea rn ing condition (D aw & Parkin , 1981; V alentine &
B ruce, 1986), o r in the intentional learn ing condition (W inograd , 1981). In
conc lusion , the sim ila rity o f the results ob ta ined w ith in ten tiona l and inc iden tal
cond itions seem s c lea rly dem onstra ted . To no tify sub jec ts o f the subsequent
m nesic test does no t have an effec t on recognition perfo rm ance .
O n the other hand , info rm ing sub jec ts about the recogn ition test could y ie ld an
effec t o f encoding instruc tions that is hard to in terp re t. W e canno t be abso lute ly
sure that subjects d id no t use the ir usua l encod ing strategy w hatever the
instruc tions g iven . In the sam e way that sub jects canno t he lp sem antica lly
iden tify ing a written w ord (Stroop effect, S troop , 1935), o r that one has difficu lty
hearing speech as sim ple sounds (Fodor, 1983), ask ing sub jec ts to m ake a
judgem ent abou t a fac ia l fea ture does not assure us tha t a judgem ent about
persona lity tra its is not a lso m ade. To ensure tha t sub jects ac tually used an holistic
encoding strategy , D effenbacher e t al. (1981) asked them to g ive an opin ion about
the d ifficu lty of m em orising the overall structu re of the face . S im ilarly , in o ther
stud ies, instead of a b inary judgem ent (e .g . likeab le vs. un likeab le), subjects w ere
asked to m ake the ir judgem ent on a 4-po int scale (B iber e t al., 1981) or on a
7-po in t scale (W ells & H ryc iw , 1984). In this w ay one can hope to have a m ore
effec tive contribution by sub jec ts. T o say w hether a nose is short o r long cou ld
hard ly ho ld the sub jec ts’ atten tion . In order to offset th is d isadvantage, Sporer
(1991) alterna ted w ithin a session the featu re on w hich judgem ents w ere m ade .
For exam ple, subjects w ere asked abou t nose size at item 1, abou t the gap betw een
the eyes a t item 2, e tc . H owever these precau tions are not en tire ly sa tisfy ing . T he
m ost direct con tro l w as ob tained by B loom and M udd (1991). They m easured
toge ther tim e inspec tion, eye m ovem ents and pup il d iam eter. T hey found tha t as
m ore eye m ovem ents w ere recorded , m ore fea tures w ere processed . H owever th is
record ing is use fu l on ly if w e consider, as B loom and M udd did, that personality
tra it judgem ents requ ire the exam ination of num erous physica l fea tures and
cannot be done at first g lance. In sum m ary , it is d ifficu lt to ob tain an equ iva lence
betw een the encoding requested and the encod ing m ade .
Tak ing into accoun t stud ies w here no judgem ents w ere m ade during
encoding , the da ta show consisten tly that m em ory for faces is invariab ly
superio r o r equally fo llow ing no judgem ents (standard inten tional encod ing)
com pared w ith fo llow ing judgem ents abou t a fac ia l featu re (see T ab le 2).
M oreover, fac ia l recogn ition is superio r o r equa l follow ing judgem ents about a
persona lity trait com pared w ith recogn ition after no judgem ents, and after a
judgem ent abou t a facial fea ture . So w e cou ld state that judgem ents abou t a
FACE RECOGNITION 549
persona lity trait o f a face lead to better recogn ition perfo rm ance than no
judgem ent (standard inten tiona l encoding) w hich in tu rn leads to a better
recogn ition perform ance than judgem ents about a physica l fea ture . T here fore ,
encod ing instructions invo lving judgem ents about a face fac ilita te (w hen a
persona lity tra it judgem ent is requested ) o r d isrup t (when a fac ial featu re
judgem ent is requested ) the recogn ition of the face .
Exposure Duration
Severa l au thors argue tha t judgem ents about persona lity tra its requ ire m ore tim e
to be ach ieved than do judgem ents about fac ia l fea tures (e.g . Badde ley &
W oodhead , 1982 ; D aw & Park in , 1981). G oldste in and Papageorge (1980)
ob tained data that seem s to invalida te such a hypothesis. Judgem ents about
beau ty and attrac tiveness of the face w ere done w ith in 150m s, w hich is m uch
shorter than the tim e allow ed to subjects to m ake judgem ents in the present
rev iew (i.e . 5 to 15 seconds). In con trast, B loom and M udd (1991, Experim en t 1)
recorded inspection tim e w here sub jec ts were asked to close the ir eyes as soon
as they had m ade the ir judgem ent. T he resu lts show ed tha t no judgem ent and
gender judgem ent d id not requ ire d iffe ren t p rocessing tim e (0 .39 s vs. 1 .09 s, no
sign ifican t d iffe rence), bu t tha t these bo th requ ired less tim e than honesty
judgem ents (2 .78 s).
T he exposure duration fac tor cou ld be confounded w ith the encoding type
facto r. In M ontgom ery ’ s study (cited in D effenbacher e t a l., 1981) the
recogn ition perform ance varied w ith encoding conditions m ade w ithin eigh t
TABLE 2Face Recognition Performance as a Function of Three Different Instructions
H its
A utho r(s ) F ea tu re S ta nd a rd P e rson a li ty
B ibe r e t a l . (1 98 1 ) .6 5 < .77 = .81
B loo m & M u dd (19 9 1) .5 6 = .48 < .76
B laney & W in o gra d (1 9 78 )a
.6 6 = .64 < .74
D e ffe n bach er e t a l . (19 8 1 ; E x p . I) .9 1 = .93 = .94
D e ffe n bach er e t a l . (19 8 1 ; E x p . II) Ð .91 < .96
L ig h t e t a l. (1 97 9 ; E x p . I)b
Ð .59 < .61
L ig h t e t a l. (1 97 9 ; E x p . II)c
Ð .67 = .67
M aye s e t a l. ( 19 80 ) .4 1 < .46 < .50
S m ith & W in o gra d (19 78 ) .7 5 = .75 < .86
S po re r (19 9 1) .5 6 < .67 = .66
W arrin g ton & A ck ro yd (19 75 )d
.8 0 = .78 < .86
D ash es in d ica te no d a ta av a ila b le ; th e d iffe re n ces a re s ig n if ic an t a t the .05 lev e l a t le a s t; < infe rio r;
> sup e rio r; = no s ig n i fic an t d iff ere n ce .aD a ta o b ta in ed w ith 5±1 0-y ea r-o ld ch i ld re n .
bD a ta o b ta in ed w ith ty p ica l fa c e s .
cD a ta ob ta in ed w i th
u n usu al fa ce s .dH its p lu s Fa lse Po s it ive s .
550 COIN AND TIBERGHIEN
seconds. On the o ther hand, no recogn ition perfo rm ance d ifference occurred
betw een encod ing conditions w hen the presenta tion of the targe t stim ulus was
repea ted . T he ce iling effect tha t cou ld appear w ith a long exposure dura tion does
not a llow clea r inte rpre tations. O nly experim en ts in w hich both encod ing
instruc tions and dura tion of study process vary cou ld allow us to d issociate the
effec t o f these two fac tors. For exam ple , D aw and Parkin (1981 ; see also Parkin
& H ayw ard, 1983 , Experim en ts 1 and 4) recorded the dec ision tim e requ ired to
m ake varied judgem ents abou t faces. The resu lts show that judgem ents abou t
persona lity tra its are m ade m ore qu ick ly than judgem ents abou t a d istinc tive
fac ial fea ture (2255m s vs. 3280m s). If w e consider the resu lts ob ta ined by
B loom and M udd (1991) and by Park in and h is collaborato rs (abou t a physical
fea ture, a persona lity tra it, and a d istinc tive fac ial fea ture ), any judgem ents
cou ld be ach ieved with in three seconds. Unfortuna te ly , in the stud ies exam ined
here , the encod ing tim e w as a lw ays greate r than five seconds, so it is no t
possib le to d issocia te the exposure dura tion fac tor and the encod ing type fac to r.
A m ore sa tisfacto ry paradigm w ould allow sub jects on ly the dura tion necessary
to m ake a judgem ent. Indeed, it cou ld be argued tha t in the facial fea ture
judgem ent cond ition, as the dura tion allow ed is g reate r than the necessary
dura tion to m ake such a judgem ent, sub jects could also have suffic ien t tim e to
m ake a personality trait judgem ent.
Exposure Duration and Judgement about aDistinctive Feature
In the d istinctive fea ture judgem ent cond ition (e .g . Reyno lds & Pezdek , 1992,
Ex perim en t 2; Valen tine & Bruce , 1986, E xperim ent 3 ; W inograd, 1978,
1981) the sub jec ts w ere asked to exam ine the fac ia l featu res and to se lec t the
fea ture tha t seem ed to be the m ost d istinctive . T he level of recogn ition
perfo rm ance ob tained w as not d ifferent from the recognition ra te ob ta ined
fo llow ing a personality tra it judgem ent. T he au thors consider tha t th is task
requ ires the ex traction of a large num ber of fea tures (e .g . W inograd , 1981).
H ow ever the recognition perform ance is ra the r varied, that is, from 68.8% to
91.2% . In all these experim en ts, the encod ing instruction fac tor is a be tw een-
subjec ts facto r. So , each sub jec t w as asked to perfo rm only one task , w hich
consisted of e ithe r m ak ing a judgem ent abou t a persona lity tra it o r choosing
the m ost distinc tive featu re . M oreover, the list from w hich the sub jects
se lec ted a featu re d id no t vary w ithin the study , and th is list con tained a
various num ber of featu res (from 5 to 11). T he sub jects knew w hich fea tures
they had to observe during the exposure of the ta rget face and it w as probab ly
easier to observe five featu res during five seconds (Courtois & M uelle r, 1979)
than nine fea tures during th ree seconds (Daw & Park in, 1981). So it can be
assum ed tha t the perfo rm ance ob tained depended in part on the exposure
dura tion of the target face (from 3 to 20 seconds), and a lso depended on the
FACE RECOGNITION 551
leng th of the list from w hich a fea ture w as selected (from 5 to 11) (see
F ig . 1 ).
INTERPRETATIONS
Depth of Processing
T he data showing that judg ing the in ferred charac ter o f a face (e.g . likeab ility)
leads to be tter recogn ition perfo rm ance than a judgem ent abou t a physical
featu re, have been inte rp re ted by o thers w ith in the leve ls o f p rocessing
fram ew ork (e .g . B ow er & K arlin , 1974 ; M cK elvie, 1985; Patterson & B adde ley ,
FIG. 1. Face re cog n i tio n p e rfo rm an ce as a fun c tio n o f ex po sure d ura t ion o f th e ta rg e t fa c e. L e tte rs
ab ov e the x -a x is re pre sen t ex p er im ents fro m w hich da ta w e re ob ta in ed .aD aw an d P a rk in (19 8 1).
bC o urto is and M u e lle r (19 79 ).
cV a len tin e an d B ru ce (1 9 86 , E x pe rim en t 3).
dPa rk in an d G o od w in
(1 9 83 ).e
W ino gra d (1 98 1 , E x p eri m en t 1 ).fW ino g ra d (1 98 1 , E x p er im ent 2).
gD effe n bach e r e t a l.
(1 9 81 ). T h e y-a x is o n the rig h t r ep re sen ts th e nu m be r of fe a tu re s fro m w hich su b jec ts ch o se a fea tu re .
W hen the lis t c o n ta in ed 1 1 fe a tu re s (E xp e rim en ts a , c , e , f, a n d g ), th e n um b e r 9 w as p re fe rre d
b ecause of the tw o irre lev an t fe a tu re s othe r and n on e .
552 COIN AND TIBERGHIEN
1977 ; W inograd, 1976). T he leve ls o f processing hypothesis w as dem onstra ted
using verbal m ateria ls (C raik & Lockhart, 1972). These authors show ed that
w ords processed sem antically w ere rem em bered be tter than w ords processed for
the ir physical charac te ristics. Bow er and K arlin (1974) w ere the first to app ly
this theoretica l no tion to m em ory for faces. The sub jec ts w ere asked to m ake
d icho tom ic judgem ents abou t gender and abou t the perceived honesty or
likeab ility of a face . B ow er and K arlin show ed tha t m em ory for faces is
considerab ly be tter fo llow ing a deep encoding (e.g . judgem ent abou t likeability)
than a superficia l encod ing (e .g . judgem ent abou t gender). T hey argue that as
p rocessing leve l increases, a larger num ber of associa tions to the targe t a re
fo rm ed in m em ory which enhances m em ory perfo rm ance .
H ow ever, th is hy po thesis has b ee n ch alleng ed b y n um e rous au thors .
W inograd (1981) a rgues tha t it is the am ount ra the r than the type of info rm ation
encoded tha t enhances the recogn ition rate (e laboration hyp othesis). Indeed ,
w hen the sub jec ts w ere asked to choose the m ost d istinctive fac ia l featu re ,
recogn ition perfo rm ance w as as high as fo llow ing a judgem ent abou t a
persona lity tra it (B laney & W inograd , 1978 ; C ourto is & M ueller, 1979; D aw &
Park in , 1981 ; D effenbacher et al., 1981 ; Park in & G oodw in, 1983; Park in &
H ayw ard, 1983 ; Sm ith & W inograd , 1978 ; Valen tine & Bruce, 1986 ; W inograd ,
1978 , 1981). M ore prec ise ly, W inograd and h is co llabora tors argue that a
persona lity trait judgem ent invo lves the encod ing of a large num ber of fea tures,
w hich enhances the probab ility o f encod ing a d istinctive fea ture . Courtois and
M ueller’ s (1979) hypo thesis, a lthough sligh tly d ifferen t, leads to the sam e
conc lusion . A ccording to these au thors, a judgem ent about a persona lity tra it
(e.g . honesty ) requires the exam ination of num erous facial fea tures in order to
a llow the m atch ing be tw een the presented face and a m em orised ``honest’ ’ face
proto type . T hus, contra ry to W inograd’ s (1981) ideas, the scann ing of num erous
fea tures is required befo re judgem ent can be m ade . O n the o ther hand , a
judgem ent about a physica l facia l fea ture requires a decision on only one
fea ture.
K err and W inograd (1982) c ritic ise the in feren tia l na ture of such a
hypo thesis. Indeed , w e cannot be sure tha t when the subjects m ake judgem ents
abou t personality they ip so fac to encode a la rge num ber of fea tures. Kerr and
W inograd found tha t face recogn ition increased w ith the num ber of desc rip tive
sentences read during the encod ing phase. Such a resu lt is in agreem ent w ith
W inograd ’ s hypo thesis. M ost in teresting ly , in a th ird exper im en t they included a
reca ll te st to be sure tha t the encod ing of the sentences w as ac tua lly done by the
subjec ts. H ere the recogn ition ra te d id not vary w ith the num ber of sentences
read during the encod ing phase. S im ila rly , the recogn ition perfo rm ance does not
vary w ith the num ber of sta tem en ts (3 vs. 15) show n w ith the targe t face
(Baddeley & W oodhead , 1982). A s suggested by an anonym ous rev iew er, an
a lte rna tive argum ent cou ld be proposed . T hat is, a longer than necessary
exposure dura tion for the ``encode a fea ture /encode a d istinc tive featu re’ ’
FACE RECOGNITION 553
in struc tion m ay a llow sub jects to encode m ore featu res than the instruction
d icta tes is necessa ry , and th is in itself m ay fac ilita te the invo lun tary judgem ent
o f personality .
If a judgem ent abou t persona lity induces the encod ing of severa l fac ial
featu res, a reconstruc tion task w ith an Identik it2
carried ou t a fte r such a
judgem ent w ould be done m ore successfu lly than fo llow ing a physica l featu re
judgem ent. U sing a be tween-sub jec ts facto ria l design, W ells and H ryc iw (1984)
asked their sub jec ts to m ake a judgem ent about a persona lity trait o r abou t a
physica l featu re, and then subm itted them to tw o diffe ren t tests: eithe r a
recogn ition test o r a reconstruc tion tes t. A lthough the recogn ition ra te ob tained
confirm s the earlier da ta , pe rform ance on the reconstruc tion task was better
fo llow ing judgem ents re lated to physical fea tures than after judgem ents about
the perceived honesty of the face . Th is resu lt calls into question W inograd’ s
e laboration hypo thesis. H ow ever w e probab ly need to c larify the cond itions in
w hich such resu lts have been ob ta ined . F i rs t, the da ta tha t cha llenged
W inograd ’ s hypo thesis cam e from reca ll tests (K err & W inograd , 1982) and
from reconstruc tion tasks (W ells & Hryc iw , 1984). It can be assum ed tha t such
tasks require a d ifferent p rocess from a recogn ition task (e .g . C om ish , 1987). In
a sim ilar way , Badde ley and W oodhead (1982) d id no t observe any d ifference in
recogn ition rate afte r the encod ing of various statem en ts p resented w ith the
ta rge t face . B ut this situa tion is rather d iffe ren t from the cond ition in which
sub jec ts a re asked to m ake judgem ents abou t physica l fea tures.
A no ther a ttem pt to test these tw o hypo theses w as carried ou t by B loom and
M udd (1991). In order to assess how m any featu res of the target w ere exam ined
in the tw o con trasted cond itions (persona lity and fea ture judgem ents) they
recorded the num ber of eye m ovem ents during the encoding phase . They
observed tha t m ore eye m ovem ents w ere m ade in the personality trait judgem ent
than in the fea ture judgem ent cond ition . H owever, although this resu lt is
com patib le w ith the e laboration hypothesis (m ore featu res are encoded during
persona lity judgem ents by m ore eye m ovem ents), it does no t ipso facto infirm
the depth of p rocessing hypothesis.
Level of Processing and Implication of the Subject
M uelle r et al. (1979) a rgued tha t a dec ision invo lving a se lf-reference criterion is
deeper than one involv ing a neutra l abso lute criterion . In order to test th is
hypothesis, sub jec ts w ere asked to base their decision on a personal criterion
(e .g . ``Is the person as ta ll as you? ’ ’ ), o r on an abso lu te criterion (e.g . ``Is the
person over 5 ft 6 in ta ll? ’ ’ ). H ow ever, no sign ifican t difference w as found
be tw een the recognition decisions m ade, w ha tever the type of encod ing
2A n Id en tik i t enab le s th e co m po s ition or the re co n stru c tion of a fa c e fro m bla ck an d w h ite
p h o to g ra ph s o f th e fiv e m a in fa c ia l f ea tu re s ; ha i r, e ye s , n o se , m o u th , and ch in .
554 COIN AND TIBERGHIEN
instruc tions (see also M ueller, C ourto is, & Bailis, 1981). It seem s that the
im plica tion of the sub ject leads to a h igher level o f recogn ition . For exam ple ,
w hen the question : ``Does this person look like you? ’ ’ is asked during the
encoding phase, the recogn ition ra te is superio r to tha t fo llow ing fea ture
judgem ents bu t does no t d iffer from those follow ing persona lity judgem ents
(Courto is & M ueller, 1979). In a ra ther sim ilar w ay , Sporer (1991) ob ta ined a
better recogn ition ra te w hen the sub jec ts them se lves generated re levan t fea tures
fo r the judgem ent abou t a physical featu re than after a judgem ent about a fea ture
g iven by the exper im en te r. H ow ever th is e ffec t w as no t observed w ith
judgem ents abou t persona lity tra its. T here fore w e m ay expec t an in terac tion
betw een the type of encod ing and the im plica tion of the subject.
Componential and Holistic Processing
N um erous au thors have interp re ted recognition perfo rm ance as a function of the
orien ting tasks w ith in the com ponential/ho listic fram ew ork . Several types of
judgem ents a re seen as be ing ho listic :
1 . Judgem ents abou t persona lity (M ayes, M eude ll, & N eary, 1980 ; Pa tte rson
& B adde ley , 1977).
2 . Judgem ents abu t a hobby or about the pro fession tha t the person depic ted
w as likely to en joy or to prac tise (K latzky , M artin , & K ane, 1982 ; Spore r,
1991).
3 . Jud gem ents ab ou t the w eigh t o f the p erson dep ic ted (W oodhead ,
Baddeley, & Sim m onds, 1979).
4 . Judgem ents abou t the global structu re of the face (D effenbacher e t a l.,
1981).
O n the o ther hand, judgem ents abou t physica l fac ia l featu res and abou t
gender are supposed to be done in an analy tica l way (Galper & C osta , 1980 ;
W ells & H ryc iw , 1984). T he com ponen tia l/ho listic d icho tom y seem s to be only
a fram ew ork to in terp ret the resu lts ob tained . Indeed , none of the stud ies has
d irec tly com pared the effects o f holistic judgem ents (e .g . scann ing the overa ll
struc tu re of the face) and com ponen tia l judgem ents (e.g . look ing a t a fac ial
fea ture) on subsequent recogn ition . D effenbacher et al. (1981) asked sub jec ts to
perfo rm an ho listic task , i.e. to scan the g loba l struc ture of a face . N o difference
w as ob tained in a yes/no recogn ition task w hatever the encod ing task (ho listic ,
standard , or judgem ents abou t a d istinctive featu re).
Judgem ents Abou t Person ality Tra it or In te lligence . If such a judgem ent
cou ld be carried out from one or tw o featu res, it cou ld be perfo rm ed analy tically .
O n the other hand , the processing of num erous fac ia l featu res cou ld be either
holistic o r ana lytic . In a review abou t soc ia l a ttribu tion, Shepherd (1989) po ints
FACE RECOGNITION 555
ou t tha t the re is no rea l ag reem ent am ong sub jec ts abou t physica l cues in
ranking faces on in te lligence . A s various and d ifferen t fea tures are considered
by subjects, it is likely tha t num erous physica l fac ia l cues a re taken in to account
instead of a few , fo r exam ple , th in eyebrow s, large or sm all eyes. A s such , w e
canno t know if the processing perfo rm ance is analy tic or ho listic in na ture .
Judgements About W eight and H eigh t. Fac ial recogn ition ra te is different
fo llow ing judgem ents about in telligence and physical featu res (W inograd, 1976)
bu t no t follow ing judgem ents abou t in te lligence and w eigh t (M ueller e t a l.,
1979; W inograd , 1976). A ccord ing to W inograd (1976), a judgem ent about
w eigh t requ ires a m ore g loba l p rocess than a judgem ent abou t a physical fea ture ,
because w eigh t is a g loba l charac teristic o f the person. One cou ld argue tha t
judgem ents abou t inte lligence and w eigh t bo th requ ire an ho listic assessm ent
(e .g . W oodhead e t a l., 1979). H ow ever, these explanation s need further
c larifica tion . F irst, we canno t d raw a conc lusion from the lack of d iffe rence
be tw een recogn ition ra te a fter judgem ents about in telligence and about w eigh t.
Second , it has been argued tha t judgem ents abou t in telligence lead to a h igh
recogn ition ra te becau se the processing w as supposed to be ho listic.
Judgements Abou t G ender. In everyday life , judgem ents abou t gender
could be based on sim ple cues (e .g . presence of a tie , m ake-up). M oreover, th is
is not possib le in the laborato ry w hen these cues are care fully discarded (e .g .
B ow er & K arlin, 1974 ; Carey e t al. 1980; M cK elv ie , 1 978 , E xperim ent 4 , 1985;
S trnad & M ueller, 1977). Such a decision could then be based on severa l local
cues (th ickness of eyebrow s, eye size) o r on m ore g loba l cues (shape of the jaw ),
o r even on re la tionsh ip betw een fac ia l featu res as suggested by O ’ T oole ,
M illw ard, and A nderson (1988). If w e consider tha t just one cue is no t su ffic ien t,
such a judgem ent could require an ho listic p rocess (e .g . B row n & Perrett, 1993).
Ind irec t da ta show tha t gender judgem ents a re processed ho listica lly . L ow
spa tia l frequenc ies convey in form ation re la ted to g loba l charac te ristics o f the
face and are integ ra ted by the v isual system earlier than h igh spa tia l frequenc ies
(B reitm eyer, 1984 ; Ericksen & Schu ltz, 1979). B u t Sergen t (1986 , E xperim ent
1 ) found tha t ca tegorisa tion based on gender is perfo rm ed faster than tha t based
on profession . M oreover , gender categorisation is im proved less by the presence
of h igh spa tia l frequencies than profession ca tegorisation . T h is indicates tha t
gender categorisation is m ain ly perfo rm ed through g loba l fac ial characte ristics
(low spa tia l frequenc ies). In terp reting such a resu lt w ith in the com ponen tia l/
ho listic fram ework , one cou ld w onder w hy gender judgem ents a lw ays lead to
low er recogn ition ra tes than personality judgem ents in yes/no recognition tasks
(B ow er & K arlin, 1974 ; S trnad & M uelle r, 1977) and in old /new recogn ition
tasks (Carey et a l., 1980 ; M cK elv ie , 1978 , 1985). Fo llow ing B addeley (1979),
one cou ld argue that th is type of encod ing is no t appropria te to distingu ish faces
of the sam e sex a t the test stage . Indeed the judgem ent m ade a t the encod ing
556 COIN AND TIBERGHIEN
stage is a judgem ent d istingu ish ing a fem ale face from a m ale face . T his k ind of
judgem ent involv ing the processing of physical featu res cou ld be qu ite useless at
the test stage in d istingu ish ing one fem ale face from the other fem ale faces seen
during encod ing .
Judgem ents A bou t a D istinc tive Fea ture . Judgem ents about a d istinc tive
fea ture, like judgem ents abou t a persona lity trait, lead to a high level o f
recogn ition. A s m entioned earlier, these resu lts a re in agreem ent w ith the
hypo thesis of W inograd (1981) which con tends tha t a judgem ent abou t a
d istinctive fea ture leads to a high leve l o f recogn ition because of the great
num ber of featu res p rocessed . H owever, an a lternative in terp reta tion could be
proposed. In on ly one study (e .g . Courtois & M uelle r, 1979) d id judgem ents
abou t a d istinc tive fea ture lead to a low er recognition perfo rm ance than
judgem ents abou t personality . In teresting ly, the list used by these au thors
con ta ined no featu re induc ing ip so facto an ho listic p rocessing , i.e . ch in, lips,
nose , eye, and forehead . On the contra ry, in the o ther stud ies the list included at
least one facia l fea ture im plica ting a global p rocessing of the face : ``head
shape ’ ’ (V a len tine & Bruce , 1986 , E xperim en t 3); ``head shape ’ ’ , and ``sk in ’ ’
(D aw & Parkin , 1981; D effenbacher e t a l., 1981 ; W inograd , 1981 , E xperim ents
1 and 2); ``ha ir’ ’ (Pa rkin & G oodw in, 1983). W e cou ld argue that because of
such fea tures the na ture of the processing is ho listic during distinc tive fea ture
judgem ents as w ell as during judgem ents abou t persona lity .
Componential/Holistic Processing and HemisphericSpecialisation
N um erous data ind icate tha t the left hem isphere is specia lised in com ponen tial
inform ation processing w hile the righ t hem isphere is specia lised in holistic
inform ation processing (e .g . B radshaw & N ettleton , 1983 ; Y oung & Ellis,
1989). G alper and C os ta (1980) exam ined the influence of ana ly tic and holistic
encoding on subsequen t fac ia l recogn ition . In the encod ing phase, targe t faces
w ere accom panied by soc ial or physica l in form ation abou t each face (e.g .
respec tive ly ``The wom an is v ivac ious and ou tgo ing’ ’ , ``T his w om an has curved
eyebrows and th in lips’ ’ ). T he test faces w ere presen ted e ither in the righ t v isual
fie ld or in the le ft v isua l field . The resu lts show im portant inte rind ividual
d ifferences and an encoding by v isua l fie ld in terac tion (see a lso Proudfoo t,
1982 , E xperim en t 2 , for sim ila r results). M oreover, th is lateral pa tte rn is no t
ob ta ined in the stan dard in ten tion a l lea rn in g parad igm (E xpe rim en t 2 ) .
Th ere fore , faces a re presum ably processed prefe ren tially by the right o f left
hem isphere accord ing to the encod ing type . T he left hem isphere recogn ition
perfo rm ance was superior to the righ t one fo llow ing fac ial fea ture judgem ents,
w hereas the righ t hem isphere w as superio r fo llo w ing judgem ents abou t
persona lity. Such a resu lt is quite in agreem ent w ith da ta show ing a left
FACE RECOGNITION 557
h em ispheric su perio rity in ana ly tic p rocessing , and a righ t hem ispher ic
super iority in holistic p rocessing (e .g . M agaro & M oss , 1989 ; V an Kleeck ,
1989). H ow ever the G alper and Costa study is no t very conv inc ing . F irst, the
la tera l exposure dura tion w as too long (300m s) to prevent possib le ocu lar
m ovem ents. G a lper and Costa (1980) po inted ou t tha t th is dura tion m akes the
resu lts m ore robust: indeed in sp ite o f th is, hem ispheric d ifferences w ere
observed . Second, one could argue that the partly verbal encoding (through the
sen ten ces ac com pan y in g the ta rg e t face ) fav ou red the le f t h em isph ere
processing (e .g . M oscov itch, Scullion , & C hristie, 1976 ; Servos & Pete rs, 1990).
Analytic Processing: A Childhood Specificity?
Som e stud ies show that ch ildren process faces analy tica lly , fea ture by fea ture ,
un til 11 or 12 years o ld and then process faces holistica lly (e.g . C arey &
D iam ond, 1977 ; F lin , 1985 ; F lin & D ziuraw iec, 1989). If ch ildren process faces
in an ana lytic w ay , standard in ten tiona l learn ing m ust no t lead to different
recogn ition perform ance from encoding abou t gender. Indeed , B laney and
W inograd (1978) found a resu lt com patib le w ith such a hypo thesis. H ow ever
th is hypo thesis cannot be m ethodo log ica lly accep ted because no conc lusion can
be infe rred from an absence of d iffe rence . On the o ther hand , if ch ildren up to 11
or 12 years o ld process faces in an analy tic w ay and are no t ab le to adop t an
ho listic strategy , the recogn ition perfo rm ance produced shou ld no t be different
according to the type of encoding . Bu t B laney and W inograd (1978) and Carey
e t a l. (1980) ob ta ined the sam e pa ttern of resu lts w ith ch ildren as tha t observed
w ith adu lts; judgem ents about persona lity led to a be tter recogn ition ra te than
judgem ents about gender. Judgem ents abou t personality cou ld allow ch ild ren to
adop t an ho listic strategy that they do no t use spon taneously . H ow ever, w ithout
a con trol g roup it is no t possib le to draw any conclusions.
T his po tential capac ity , if it ex ists (see fo r exam ple B aenn iger, 1994), does
no t seem to vary w ith age. Indeed the m agn itude of the im provem ent in
recogn ition allow ed by persona lity encoding com pared w ith the recogn ition
leve l ob ta ined after featu re encoding, does not seem to vary w ith the age of the
sub jec ts (from 7 to 14 years o ld; Carey e t a l., 1980). T hus, as child ren can
benefit from instruc tions to encode faces, the ana lytic p rocessing canno t be
considered as a childhood spec ific ity . M oreover, recen t investigations show tha t
the effic iency of bo th featu ral encod ing and ho listic encod ing increase w ith age
(C hung & T hom son , 1995 ; E llis & E llis, 1994; Pac teau & B on thoux , 1994).
T he dep th of p rocessing m ode l and the e labora tion m odel a re d ifficult to test.
Superio r recogn ition perfo rm ance m ay be assum ed to arise from deeper
p rocessing, and this requ ires a m ore e labora te degree of fea ture encoding .
M oreover, as m en tioned earlie r, we canno t be sure that the subject ac ts in
accordance w ith instruc tions. H ow ever, the com ponen tial/ho listic fram ew ork
558 COIN AND TIBERGHIEN
offers a powerfu l w ay of exam in ing the resu lts. It w ould be possib le to
m anipu late the stim uli in order to induce a com ponentia l o r a ho listic p rocessing
of the face . M ore prec isely , w e cou ld ask subjects to m ake judgem ents abou t
persona lity o r abou t a fac ia l fea ture, and to m anipula te the test face (p resen ted
during the recogn ition test) in order to con tro l the type of processing carried ou t.
For instance , the test face cou ld be presen ted upside dow n, so in terfer ing w ith
configura tional (and ho listic ) p rocessing , o r it cou ld be b lurred , so in terfe ring
w ith com ponential p rocessing . T he test face cou ld a lso be filtered, either
con ta in ing h igh-pass spa tia l frequenc ies (conveying fine de ta ils) o r low -pass
spatia l frequenc ies (convey ing coarse info rm ation) (see C oin , 1993 for a
presen ta tion of the spa tia l frequency approach on face recognition ).
IMPROVEMENT OF FACE RECOGNITION
Type of Processing Considered Most Effective bySubjects
B efore exam in ing the possible im provem ent o f face recogn ition , one cou ld ask
subjec ts them se lves. Irrespec tive of the ir e ffec tive w ay of encod ing faces, w hich
is the m ost e ffec tive strategy , as they see it? To look care fu lly at a fac ia l fea ture
cou ld im prove face recogn ition. Th is asse rtion , considered as convey ing the
com m onsense v iew (C ourto is & M ueller, 1979), seem s to be confirm ed in an
inform al w ay by W inograd (1978 , 1981 ; see also M ueller & W herry , 1980).
Indeed , W inograd usually asked the in troduc tory psycho logy studen ts to guess
w hich strategy is m ost effec tive ; judg ing faces in term s of persona lity , judg ing
faces in te rm s of a fac ia l fea ture , or encod ing faces w ithou t any judgem ents (that
is to say in the ir ow n way). Invariably , the m ajority of the studen ts vo ted fo r the
fea ture processing stra tegy . B addeley and W oodhead (1983) first separa ted
subjec ts in to tw o groups according to the recogn ition ra te ob ta ined in a prev ious
experim ent. T he first g roup ob tained a higher leve l o f recognition than the
second group . T hen three recogn ition tests w ere adm in istered, am ong them a
fac ial recogn ition test. The resu lts ob ta ined w ere qu ite in agreem ent w ith the
two-c lass separation carried ou t: the first g roup obta ined significan tly superior
recogn ition perfo rm ance to the second group . H owever the first g roup subjects
c la im ed tha t they had analysed faces featu re by fea ture . T his resu lt con trasts
w ith the da ta presen ted in th is rev iew w hich c lea rly show that judgem ents about
persona lity lead to the best recogn ition rate . Indeed , the resu lts o f a post-
experim enta l questionna ire (Sporer, 1991) ind ica ted that judgem ents about
persona lity, just as judgem ents abou t the m ost d istinc tive fea ture , are considered
to be m ore effec tive than judgem ents abou t physical fea tures. A re sub jects
consc ious of the m ost effective stra tegy for m em orising faces? The agreem ent
betw een w hich stra tegy is considered to be the m ost e ffec tive by sub jec ts, and
the stra tegy that is dem onstrated experim en ta lly to be the m ost e ffec tive ,
FACE RECOGNITION 559
a lthough som etim es observed (Courto is & M uelle r, 1979 ; Spore r, 1991), cou ld
no t be considered as com ple te .
Improvement of Face Recognition
Is it possible to im prove face recogn ition perform ance? In the several stud ies
rev iew ed here , it does no t seem tha t a ceiling effect occurs. Indeed , ask ing
sub jec ts to m ake persona lity judgem ents abou t faces im proves recogn ition
p erfo rm anc e , com pared to the reco gn it io n ra te o b ta ine d af te r s tan dard
instruc tions (cf. T ab le 2). On the o ther hand, a lthough the hum an ability to
recogn ise unknow n faces is qu ite rem arkab le in laborato ry cond itions and in
everyday life , th is ability does no t alw ays reach its u ltim ate leve l. For exam ple ,
th is leve l inc reases w ith repea ted disp lay ing of faces (E llis, Shepherd , G ib ling ,
& Shepherd , 1988) or w ith an increase of exposure dura tion (M cK elv ie, 1990 ,
E xperim en t 1; R eyno lds & Pezdek , 1992 , Experim en t 1 ). So , w hich type of
train ing could perm it an im provem ent o f sub jects’ facial m nesic perfo rm ance?
W oodhead et a l. (1979) subm itted sub jec ts to a three-day tra ining course
includ ing lec tures, d iscussions, and exerc ises w ith Penry’ s (1971) Photofit3. T he
train ing consisted of observation (of stance, faces, accesso ries from photo-
g raphs), m em ory (orien ted tow ard s featu re se lec tion from the P hotofit k it), and
verba l desc rip tion (also derived from the P hoto fit). The resu lts ob tained are
ra ther puzz ling : no im provem ent (in a recogn ition task : Experim en t 1 , o r in a
m atching task: Experim en t 2 ), bu t ra the r a decrease of perform ance (in a
m atching task : E xperim ent 3 ) w as observed . A sim ilar resu lt w as ob ta ined by
M alpass, L av igueur, and W eldon (1973 , E xperim ent 1 ) w ith verbal tra in ing .
T he argum ents p roposed to exp lain such lack of im provem ent are as follows:
1 . T his nega tive resu lt could be ascribed to the use of the Photofit which
induces a fea tura l p rocessing of the face, tha t is, o f the eye , the nose , etc .
(W oodhead et al., 1979);
2 . T he process of recogn ising faces is so overlea rned through life tha t any
ex tra tra in ing cou ld not sign ifican tly im prove it (B adde ley & W oodhead , 1983;
D effenbacher, 1988 ; W oodhead et a l., 1979);
3 . T he learn ing m ethods se lec ted are no t su itab le and too artific ial (M alpass,
1981; E llis, Shepherd , & D avies , 1975).
It is no t surp ris ing tha t a train ing procedure favouring the ana lytic p rocessing
of a face (e.g . W oodhead et a l., 1979) does no t perm it an im provem ent of
m em orisa tion . Such a stra tegy , induced by encod ing instructions, leads to a
low er m nesic perform ance than any o ther type of encod ing instruction (standard
3A P ho to fi t en ab le s th e co m po s it io n or re con s tru c tion o f a fa c e by sup e rim po s it io n o f
t ra nsp ar en ci e s, on w h ich th e m a in fa c ial fe atu re s a re re pre sen ted .
560 COIN AND TIBERGHIEN
lea rn ing , judgem ents abou t personality or abou t a d istinctive fea ture ). One cou ld
suspec t tha t verba l tra in ing (e .g . M alpass e t al., 1973) does no t im prove face
m em orisa tion e ithe r. Indeed , the quality o f the verba l desc rip tions does no t
corre late w ith the capac ity for face iden tifica tion (D effenbacher, 1988; P igot t &
B righam , 1985). Verbalisa tion of faces has been show n to have on ly a few
positive effects on recogn ition perfo rm ance (C hristie & E llis, 1981 ; K latzky &
Forrest, 1984, Experim en t 3 ), and even a nega tive effect (Schoo ler & E ngstler-
Schoo ler, 1990). Subjects them se lves judge verba l desc rip tion of a face to be
m ore d ifficu lt than m em orisa tion (L aughery , D uval, & W ogalte r, 1986).
M oreover, to verba lise on ly sligh tly im proves recognition perform ance (B ruyer,
1982). H owever, ve rba l descrip tions of faces generally lead to a low er m em ory
perfo rm ance than reconstruc tion techn iques using Photofit and Identik it for
c rim ina l investiga tions (e.g . D av ies, 1983), w hich in turn lead to a low er
m em ory perfo rm ance than recogn ition tasks (see e .g . T ibergh ien , 1983 , fo r a
desc rip tion).
A s m entioned earlier, judgem ents about persona lity traits lead to h igher
recogn ition perfo rm ance than standard learn ing instruc tions. W ould such a
stra tegy within a tra ining course perm it an im provem ent o f face recogn ition?
M alpass (1981) subm itted sub jects to various train ing sessions. The subjec ts o f
the first g roup w ere tra ined to ana lyse facial featu res and the ir various changes
through Penry ’ s (1971) Photofit . T he second group of subjects w ere asked to
rank faces accord ing to perso na lity characteristics. T he th ird group of sub jects
w ere asked to look care fully a t the general physical appearance of faces, and
then to rank them accord ing to the ir sim ilarity w ith four o ther faces. Four
successive recogn ition tasks w ere adm in istered to the fourth group . N o
difference to a subsequent yes/no recogn ition test occurred w hatever the
prev ious train ing .
It is possib le, as M alpass (1981) po inted ou t, tha t the sub jec ts’ ab ility in
encoding and recogn ising faces is so high that any train ing m ethod to im prove it
has no effec t. M oreover the m ethods them se lves cou ld be too crude , in v iew of
natu ral stra teg ies deve loped by subjects. F ina lly , we could argue that these
courses a re perhaps based on inadequa te hypo theses of how people process
faces. T he M alpass (1981) experim en t shows tha t tra in ing based on holistic
p rocessing of the face does no t im prove recogn ition perfo rm ance . O ne could
poin t ou t tha t som e people cou ld spontaneously process faces ho listically .
M ueller and Thom pson (1986 , E xperim en t 1 ; see a lso M ueller & Thom pson,
1988) asked sub jec ts to give as m any personality tra its as possible abou t figures,
e ither befo re the study phase or a fter the test phase . Fa lse alarm s w ere few er for
subjec ts w ho gave num erous persona lity tra its than fo r o ther sub jects. H ow ever,
correct dec isions for the fo rm er sub jec ts varied be tw een o ld and new faces. It is
possib le tha t these sub jects adop t a h igh criterion w hich enhances perfo rm ance
on new faces and reduces perfo rm ance on o ld faces. Bu t w e canno t c la im that
the recogn ition ra te is higher fo r sub jec ts who process faces ho listica lly (as was
FACE RECOGNITION 561
supposed in this experim ent) than for o thers. Th is superiority , if it ex ists, seem s
to be observed on ly in recogn ition tasks. W hen sub jec ts were asked to recogn ise
faces from Iden tik it , the ana ly tic strategy tha t consists o f exam ining each fac ial
featu re led to a h igher recognition level than the ho listic stra tegy (e.g . L aughery
e t al., 1986).
O n the other hand , W ells and Turtle (1988) noticed tha t judgem ents about
persona lity tra its lead to be tte r recogn ition perfo rm ance than do physical
judgem ents. In contras t, judgem ents about physica l fea tures a re in turn m ore
effic ien t for verba l descrip tion (or fo r reconstruction w ith an Identik it: e.g . W ells
& Hryc iw , 1984) of the studied face (see F ig. 2).
T he data presented here suggest tha t:
FIG. 2. R ecog ni tion and v e rba l de sc rip tion of fa ce s a s a fun c tio n of enco d ing in s tru c tion s . F ro m
``W h at is th e be s t w ay to en co d e fa c es ? ’ ’ by G .L . W e lls & J .W . T urt le , 1 9 88 , In M .M . G ru n eb e rg ,
P .E . M o rris , & R .N . Sy ke s (E d s .), P ra c tica l a sp ec ts o f m em o ry : C urren t re search a n d issu e s: V o l . I.
M em ory in e v er yd ay li fe , (p .1 67 ). C h iche s te r, U K : W iley .
562 COIN AND TIBERGHIEN
1 . If subjects a re asked to recognise faces am ong o thers (w ith in a crow d or in
an iden tification parade fo r c rim ina l investigation, the m ost e fficient encod ing
stra tegy is to judge faces according to in ferred characte ristics (e.g . pe rsonality
tra it).
2 . W hen the test invo lves a descrip tion of the face (fo r exam ple from an
Iden tikit or a Photofit) the m ost e ffec tive encod ing strategy consists of judg ing
faces accord ing to physica l featu res.
CONCLUSIONS
It is we ll established that encod ing instruc tions influence subsequen t fac ial
recogn ition. H ow ever the explanations proposed are con troversial. T he dep th of
p rocessing theory offe rs an a ttractive fram ew ork to expla in the resu lts obta ined .
N everthe less it has been challenged by the W inograd (1981) e labora tion
hypo thesis. U nfortuna te ly the availab le data do no t enab le us to say exac tly
w hether it is the dep th of p rocessing , o r the num ber of fea tures encoded
(elabora tion hypo thesis) that leads to be tter recogn ition perfo rm ance than the
process of one fac ia l featu re . R esu lts show ing tha t judgem ents abou t personality
perm it a be tter subsequen t recogn ition rate than judgem ents abou t physical
fea tures have been obta ined in various situa tions. A s severa l au thors have
c la im ed , it seem s that the form er would be ho listic in na tu re, and the la tter
ana ly tic . Furthe rm ore, such a hypo thesis is consisten t w ith neurophysio logical
data w hich show tha t the con tribu tion of low spa tial frequencies convey ing
holistic aspec ts o f the stim ulus is superior to the con tribu tion of high spatial
frequenc ies transm itting fea tura l info rm ation for face recogn ition (see Coin,
1993 , for a rev iew ). Unfortunate ly the com ponen tial/ho listic approach was on ly
considered as a fram ew ork to in terp ret the results ob tained . N everthe less th is
approach offers tw o m ain advan tages in eva lua ting the effec t o f encod ing
instruc tions on face recogn it ion . F irst, it is possib le to m anipu late the stim uli in
order to con tro l the in form ation ava ilab le to be eithe r ho listic by b lu rring (o r by
a low -pass spa tia l frequency filtering), o r com ponen tia l by presen ting the face
upside dow n (or by a h igh-pass spa tial frequency filtering) . Thus, w e cou ld
obta in an equ iva lence betw een the type of encoding requested and the type of
encoding perform ed , w hich is d ifficult w ithin the dep th of p rocessing and the
e labora tion m odels.
Judgem ents m ade during encod ing cou ld m ake the re trieval o f fac ial
inform ation sto red in m em ory easier. H ow ever, th is fac ilita tion depends bo th on
the type of judgem ent m ade and on re trieva l cond itions. For exam ple , a
judgem ent abou t a fac ia l featu re is m ore effec tive than a personality judgem ent
in fac ial desc rip tion or reconstruction tasks. Re lative to standard in ten tional
lea rn ing cond itions, to ask subjects to m ake judgem ents about a physical fac ial
fea ture disrupts the subsequen t recogn ition ra te. O n the con tra ry, to ask sub jects
to m ake judgem ents abou t persona lity enhances the recogn ition ra te . F rom these
FACE RECOGNITION 563
resu lts, is it possible to in fer tha t facial recognition ab ility, in spite o f its h igh
degree of e fficiency , cou ld be im proved? H ow ever, the train ing courses in tended
to im prove th is ability d id no t show convinc ing resu lts. T he fa ilu re of the
train ing procedure is p robab ly due to the artific ial situa tion of face m em orisation
in the laborato ry . O ne could assum e, on the o ther hand , tha t several fac tors a re
involved in face m em ory in everyday life , depend ing on the situation . For
exam ple , the re liab ility of eyew itness identifica tion cou ld be im proved by
contex t reinstatem en t p rocedures (e.g . Cu tler, Penrod , & M artens, 1987)
a lthough contex t cues cou ld also have an im portan t ro le in labora tory research
(e .g . D av ies, 1988; T ibergh ien , 1986).
F inally, g iven tha t accuracy is usua lly so h igh , any im provem ent m ight have
been m asked by a ceiling effec t (B adde ley , 1979). H ow ever, fac ia l recogn ition
ab ility cou ld probab ly be increased by im prov ing recogn ition la tency. Som e
stud ies have show n that low spa tia l frequenc ies are in teg rated faster than h igh
frequenc ies by the v isua l system (Bre itm eyer, 1984; E ricksen & Schu ltz , 1979).
M oreover, h igh frequenc ies o f a stim ulus could be reduced by b link ing . T hus to
b link or to screw up one ’ s eyes cou ld be an easy w ay to favour the low spa tia l
frequency processing (convey ing ho listic inform ation and in teg rated faster by
the v isual system ) a llow ing an ho listic and fast p rocessing of a face .
M an usc rip t re c e iv ed N o vem b er 1 99 5
M an usc rip t a c c ep ted 31 Ju ly 1 99 6
REFERENCES
B ad d el ey , A .D . (1 97 9 ). A p pl ied co g ni tiv e and cog ni tiv e app lied p sy ch o log y : T he ca se o f fa c e
re co gn ition . In L .-G . N i lsso n (E d .), P erspec tiv es on m em ory re sea rch: E ssa y s in h o no r o f
U p psa la U n ive rsi ty ’ s 5 0 0 th an n ive rsary (p p . 3 6 7±38 8 ). H i llsd a le , N J : L aw re n ce E rlbau m
A ssoc ia te s In c .
B ad d el ey , A .D ., & W o od h ead , M .M . (1 9 82 ). D ep th of pro ce ss ing , con tex t, a n d fa c e re cog n i tion .
C a n ad ia n Jou rna l o f P sy ch o log y , 36 , 1 48 ±1 6 4 .
B ad d el ey , A .D ., & W o od h ead , M .M . (19 8 3). Im prov in g fa c e re cog n i tio n ab ili ty . In S .M .A . L loy d-
B o s to ck & B .R . C liff o rd (E d s .), E va lu a t in g w itne ss e viden ce (p p .12 5 ±1 36 ). N ew Y o rk : W iley .
B aen n ing e r, M .-A . (19 9 4). T h e d ev elop m ent o f fa c e r ec o gn it io n : F ea tu ra l o r co nfigu rat io na l
p roce ss in g? Jou rna l o f E xp e r im en ta l C h ild P sych o logy , 4 7 , 3 77 ±3 9 6 .
B ah ric k , H .P ., B ahr ick , P .O ., & W itt lin g er , R .P . (1 97 5) . F i fty yea rs of m em o ry fo r nam es and fa ce s :
A c ro ss-s e c tio na l ap p roach . Jo u rn a l o f E xp er im en ta l P sy cho log y : G ene ra l, 1 04 , 54 ±75 .
B ibe r, C ., B u tte rs , N ., R o sen , J ., G e rs tm an , L ., & M a tt is , S . (19 81 ). E n cod in g s tra teg ies an d
re co gn ition o f fa c e s b y a lco h o lic K o rsa k o ff an d o th e r b ra in -d am aged pa tie n ts. Jo urn a l o f C l in ica l
N eu rop syc ho lo g y , 3 , 31 5±3 30 .
B laney , R .L ., & W in og ra d , E . (1 97 8 ). D ev elop m en ta l d iffe re n ce s in ch i ld re n ’ s re co gn it io n m em o ry
fo r fa c es . D eve lo pm en tal P sy cho log y , 1 4 , 44 1±4 42 .
B loo m , L .C ., & M ud d , S .A . (19 91 ). D ep th o f pro ce ssing ap pro ach to fa c e re co g n i tio n : A te s t o f
tw o th eori es . Jo urn a l o f E xp e r im enta l P sy ch o lo g y: Lea rn ing , M em o ry, a nd C o gn ition , 17 ,
5 56 ±56 5 .
B o w e r, G .H ., & K a rlin , M .B . (19 7 4). D epth of p ro ce ss in g p ic tu re s o f fa c es an d re cog n ition
m em o ry . Jo urn a l o f E xpe r im en ta l P sy ch o lo gy , 10 3 , 7 51 ±75 7 .
564 COIN AND TIBERGHIEN
B ra d sh aw , J .L ., & N e ttle ton , N .C . (1 98 3 ). H u m an ce reb ra l a sym m e try . E ng lew oo d C l iffs , N J :
P re n tice -H a ll.
B re itm eye r, B .G . (19 8 4). V isu a l m a sk in g: A n in tegra tive a p pro ach . N ew Y o rk : O xfo rd U nive rs ity
P re ss .
B ro w n, E ., & P e rre tt, D .I. (1 9 93 ). W h at g ive s a fa ce i ts gen de r? P ercep t ion , 2 2 , 82 9±8 40 .
B ru ye r, R . (1 98 2 ). R oÃle d u lan gu age e t d e la m eÂm o ire v isu e lle dan s la p e rc ep t io n de s v isag e s : E ffe t
de s leÂsion s c eÂreÂb ra le s u n i la teÂra le s [R o le o f lang uag e and v isua l m em o ry on fa c e p e rc ep tion :
E ffe c t o f u n ila te ra l b rai n in ju ri es ]. P sych o lo g ie F ra ncË a ise , 27 , 1 4 6±15 7 .
C a re y , S ., & D iam o nd , R . (19 77 ). F ro m p iec em eal to co n fig ura tio na l re p re sen ta tion o f f ac e s .
Sc ience , 1 95 , 3 12 ±3 1 3 .
C a re y , S ., D iam on d , R ., & W o o ds , B . (19 8 0). D eve lo p m en t of fa ce re cog n ition . A m atura t io na l
com p on en t? D eve lo pm enta l P sy cho lo gy , 1 6 , 2 5 7±26 9 .
C h ris tie, D .F .M ., & E l lis , H .D . (1 9 81 ). P ho to -fit co n st ruc t io ns ve rsu s v er ba l d e scri p t io ns of f ac e s .
Jou rna l o f A p p lied P sycho lo gy , 6 6 , 3 5 8±36 3 .
C h un g , M .-S ., & T h om so n , D .M . (19 9 5). D eve lo pm en t o f f ac e rec o g n it io n . B rit ish Jo u rn a l o f
P sy cho lo g y , 8 6 , 5 5±87 .
C li ffo rd , B .R ., & P r ior , D . (19 80 ). L eve ls of p ro ce ss in g an d capac ity a lloc a tion . P ercep tu a l a n d
M o to r Sk i lls , 5 0 , 82 9 ±8 30 .
C o in , C . (19 9 3). Tra item en t co m po n en tie l e t co n f ig ura l d e s v isag e s n on fa m ilier s [C om p o nen tia l
and co n fig ura l p ro c e ss ing o f u nk n ow n fa ce s]. U n pu bl ish ed d oc to ra l d isse rta t ion , G re n ob le
U nive rs ity , F ra nce .
C o m ish , S .E . (1 98 7 ). R eco gn it io n of fa c ia l s tim u li f o l low in g an in te rven in g ta sk inv o lv ing the
Iden ti -k i t. Jou rna l o f A pp lied P sycho lo gy , 7 2 , 48 8±4 91 .
C o urto is, M .R ., & M u e lle r, J.H . (1 9 79 ). P roce ss in g m u lt ip le ph y sic al fe a tu re s in fa c ia l re co gn it io n .
B u llet in o f the P sych on om ic S oc iet y , 14 , 7 4 ±7 6 .
C ra ik , F .I.M ., & L o ckh a rt, R .S . (1 97 2 ). L eve ls o f p roce ss in g : A fra m ew ork fo r m em ory re se ar ch .
Jou rna l o f V e rb a l Learn ing an d V e rba l B eh a v ior , 1 1 , 67 1±6 84 .
C u tle r, B .L ., P en rod , S .D ., & M arte n s , T .K . (1 98 7) . Im p rov in g the re lia b il ity of ey ew itn e ss
id en t ific a tion : Pu tt in g con tex t in to con tex t. Jou rna l o f A pp lied P sycho lo gy , 7 2 , 62 9±6 37 .
D avie s, G .M . (1 9 83 ). F ore n sic fa c e re c a ll: T h e ro le o f v isu a l an d ve rb a l in fo rm a tio n . In S .M .A . L loy d-
B o stock & B .R . C l iffo rd (E ds .), E valua ting w itn e ss e v id ence (p p .10 3 ±12 3) . N ew Y o rk : W iley .
D avie s, G .M . (1 9 88 ). F aces an d p lac es : L ab or ato ry re se a rc h o n co n tex t an d fa c e re co gn it ion . In
G .M . D av ie s & D .M . T h om son (E d s .), M em o ry in co n tex t: C on tex t in m em o ry (p p .35 ±53). N ew
Y ork : W iley .
D aw , P .S ., & Pa rk in , A .J. (1 98 1 ). O b se rv a tions on the e ffic ien cy of tw o d iffe re n t p ro ce ss in g
s tra teg ie s fo r re m em b erin g f ac e s . C a na d ia n Jo urn a l o f P sych o lo gy , 35 , 3 51 ±35 5 .
D e ffe nb ache r, K .A . (1 9 88 ). E yew i tn es s re se a rch : T h e nex t ten yea rs . I n M .M . G ru n eb e rg , P .E .
M o rris , & R .N . Sy ke s (E d s .), P rac tic a l a spec ts o f m em ory : C u rren t re sea rch a nd issu es : V o l . I.
M em o ry in e v e ryda y l ife (p p .20 ±26). C h iche s te r, U K : W iley .
D e ffe nb ache r, K .A ., L eu , J .R ., & B ro w n , E .L . (1 98 1) . M em o ry fo r fa c es : T e s tin g m e th od , en cod in g
s tra teg y , an d co nf iden ce . A m erican Jou rna l o f P sy ch o lo gy , 9 4 , 13 ±26 .
D evine , P .G ., & M a lp a ss , R .S . (1 9 85 ). O ri en t in g s tra teg ie s in d iffe re n tia l fa c e re co gn ition .
P erson a lity a nd S oc ia l P sy ch o lo g y B ul le tin , 1 1 , 33 ±4 0 .
E l lis , H .D ., & E llis, D .M . (1 9 94 ). D eÂv elop p em en t d e s c apac iteÂs aÁ tra ite r d e s tra n sfo rm a tio ns
fa c ia le s ch ez l’ e n f an t [D ev elop m en t o f ch i ld re n ’ s ab i lity to re cog n ize tra n sfo rm ed fa c e s].
P sycho lo g ie F ran cË a ise , 3 9 , 28 7±3 00 .
E l lis , H .D ., Sh eph e rd , J .W ., & D av ie s , G .M . (19 7 5). A n in ve s tiga tio n o f th e u se of th e Ph o tof it
te ch n iq u e fo r re c a lling fac e s . B ritish Jou rna l o f P sy cho lo gy , 6 6 , 29 ±37 .
E l lis , H .D ., Sh eph e rd , J .W ., G ibl in g , F ., & Sh eph e rd , J . ( 19 88 ). S tim ulus fa c to rs in f ac e le a rn ing . In
M .M . G ru neb e rg , P .E . M orri s, & R .N . S yk e s (E ds .), P rac t ica l a spec ts o f m em o ry : C u rren t
res ea rch a nd issue s: V o l. I. M em o ry in e v e ryd a y l ife (p p .1 3 6±14 4 ). C hiche s te r, U K : W iley .
FACE RECOGNITION 565
E ric k sen , C .W ., & S chu ltz , D .W . (1 9 79 ). I n fo rm a tio n p roce ss ing in v isua l se a rc h : A co n tin u ou s
flow con cep t ion and exp e rim ental re su l ts . P ercep tion an d P sych o ph y s ic s , 2 5 , 24 9 ±2 63 .
F lin , R .H . (19 85 ). D ev e lo pm en t o f f ac e re co gn it io n : A n en cod in g sw itch ? B rit ish Jou rna l o f
P sych o logy , 76 , 1 23 ±1 3 4 .
F lin , R .H ., & D z iu ra w iec , S . (1 9 89 ). D ev e lo pm ental fa c to rs in fa c e pro ce ss ing . In A .W . Y o un g
& H .D . E llis (E d s.), H a nd b oo k o f re sea rch on fa ce p ro ce ssing (p p .33 5±3 78 ). A m ste rd am :
E lsev ier.
F od or , J .A . (1 98 3 ). T he m od u la ri ty o f m ind : A n e ssay o n facu lty p sych o lo gy . C am b rid g e, M A : M IT
P res s .
G a lp e r, R .E ., & C o sta , L . (1 98 0 ). H em isp he ric su pe rio rity fo r re co gn iz in g f ac e s d ep en ds up o n h o w
they a re lea rn ed . C ortex , 1 6 , 21 ±38 .
G ao nac ’ h , D ., & G ib o in , A . (1 97 7 ). L a re co nn a issance d e p o rtra its: N iv eau e t sub jec tiv ite d u
trai tem ent [F igu re r ec o gn it ion : L ev e l o f pro ce ss ing an d su b jec tiv ity ]. P sycho lo g ie e t P eÂd ag o gie ,
9 , 2 5±32 .
G o ld s te in , A .G ., & P ap ag eo rg e , J . ( 19 80 ). Ju d gm en ts of fa c ia l a t tra c tiv ene ss in the ab sen ce o f ey e
m o vem en ts . B u lle t in o f th e P sycho n om ic S oc ie ty , 15 , 2 69 ±27 0 .
K e rr, N .H ., & W in o gra d , E . (19 8 2). E ffe c ts o f co n tex tua l e lab ora t io n on fa c e re co gn it ion . M em o ry
& C o gn it io n , 1 0 , 60 3±6 09 .
K lat zky , R .L ., & F orr es t , F .H . (1 9 84 ). R eco gn iz ing fa m ilia r and u n fa m iliar fa ce s . M em o ry &
C o g nit io n , 1 2 , 60 ±70 .
K lat zky , R .L ., M a rt in , G .L ., & K ane , R .A . (19 82 ). S em ant ic in te rp re ta tio n e ffe c ts on m em o ry fo r
fa c e s . M em o ry & C o gn it io n , 1 0 , 19 5 ±2 0 6 .
L aug he ry , K .R ., D u v a l, C ., & W o ga l te r, M .S . (19 8 6). D yn am ic s of fa c ia l re c a ll . In H .D . E l lis , M .A .
Je eve s , F . N ew co m b e, & A .W . Y ou ng (E d s .), A spec ts o f fa c e pro ces s in g (pp .3 73 ±38 7).
D o rd re ch t, T h e N e the rl and s : N i jh off.
L ig h t , L .L ., K ayra -S tu a rt, F ., & H olland e r, S . (19 7 9). R ecog ni tio n m em ory for ty p ic a l an d u nu sua l
fa c e s . Jou rna l o f E xp e r im enta l P sych o lo gy : H u m an Learn in g a nd M em o ry , 5 , 2 1 2±22 8 .
M aga ro , P .A ., & M o ss , B .F . (1 98 9 ). T he e ffe c t o f ana ly t ic ve rsu s h o l is tic en cod in g ins tru c tio n s o n
h em isph e ri c sup e rio rity . C o rte x , 25 , 3 17 ±3 2 4 .
M a lp as s , R .S . (1 98 1 ). T ra in in g in fa c e re co g n i tio n . In G .M . D avie s , H .D . E l lis , & J.W . S hep he rd
(E ds .), P ercei vi ng an d rem em ber in g fa c e s ( pp . 2 71 ±2 8 5). L on do n : A cad em ic P re ss .
M a lp as s , R .S ., L av igu eur, H ., & W eldo n , D .E . (1 97 3). V e rba l an d v isu a l tra in ing in fa ce
re co gn ition . P ercep t ion an d P sychop h y sic s, 14 , 2 85 ±2 9 2 .
M aye s , A ., M eu de ll , P ., & N ea ry , D . (1 98 0) . D o am nes ic s ado p t in ef fic ien t enco d ing st ra teg ie s w i th
fa c e s an d ra nd om shap e s? N eu rop syc h o lo g ia , 18 , 5 27 ±54 0 .
M cK e lv ie , S .J. (1 9 78 ). S ex d i ffe re nce s in f ac ia l m em ory . In M .M . G ru neb er g , P .E . M orr is , & R .N .
S yk e s (E ds .), P rac t ica l asp ec ts o f m em ory (pp .2 63 ±26 9) . N ew Y ork : A cad em ic P re ss .
M cK e lv ie , S .J . (19 85 ). E ffe c t o f dep th of p ro ce ss in g o n re co gn ition m em ory fo r n orm a l an d
inv e rte d p ho to g rap h s o f fa c es . P ercep tu a l a n d M otor S ki lls , 6 0 , 50 3 ±5 08 .
M cK e lv ie , S .J . (19 9 0). E ff ec ts o f ex po sure tim e an d in ve rs io n o n the co nfiden ce±accura cy
re la t io nsh ip in fa c ia l m em o ry . A te s t o f the op tim a l ity h yp othe s is . P ercep tu a l an d M o to r S k ills,
7 1 , 32 ±3 4 .
M cK e lv ie , S .J . (19 9 1). E ffe c ts o f pro ce ss in g st ra tegy and tra n sfo rm a tio n o n re co gn it io n m em ory fo r
p ho to g rap hs o f fa c es . B ul le tin o f th e P sycho no m ic So c ie ty , 2 9 , 98 ±10 0 .
M osco v i tch , M ., S cu ll ion , D ., & C hri st ie , D . (1 97 6). E a rly ve rsus la te s tag e s of pro ce ssing and the ir
re la t io n to fun c tio na l hem isph e ric a sy m m e trie s in fa c e re co g n i tio n . Jou rna l o f E xpe r im en ta l
P sych o logy : H u m an P ercep tion an d P er form a nce , 2 , 40 1 ±4 16 .
M ue l le r, J .H ., B a i lis , K .L ., & G o ld s te in , A .G . (1 97 9). D epth of pro ce ss ing and an x ie ty in fa c ia l
re co gn ition . B ri tish Jo urn a l o f P sy ch o lo gy , 7 0 , 5 1 1±51 5 .
M ue l le r, J .H ., C a rlom u sto , M ., & G o ld ste in , A .G . (1 9 78 ). O ri en ting ta sk an d s tu d y tim e in fa c ia l
re co gn ition . B u lle tin o f the P sych on o m ic S oc ie ty , 11 , 3 13 ±31 6 .
566 COIN AND TIBERGHIEN
M u e ller, J.H ., C o u rto is , M .R ., & B a i lis , K .L . (19 81 ). S e lf-re fe re n ce in fa c ia l re cog n i tion . B ul le tin o f
th e P sycho no m ic So c ie ty , 1 7 , 85 ±88 .
M u e ller, J .H ., & T h om p son , W .B . (19 8 6). S te reo ty p ing an d fa c e m em ory . I n H .D . E ll is , M .A .
Jee v e s , F . N ew com b e , & A .W . Y o un g (E d s .), A spec ts o f fac e pro ce ss in g ( pp .1 63 ±1 69 ).
D ord re ch t , T h e N ethe rl an ds : N i jh off.
M u e ller, J .H ., & T h o m p son , W .B . (1 98 8 ). In d iv idu a l d i ffe re nce s in fa ce m em ory . I n M .M .
G run ebe rg , P .E . M orri s, & R .N . S yk e s (E d s .), P rac t ica l a spec ts o f m em o ry : C urren t re sea rch
an d issu e s: V o l. I. M em o ry in e v e ryd a y li fe (p p .71 ±7 6). C hiche s te r, U K : W iley .
M u e ller, J .H ., & W he rry , K .L . (1 98 0). O rie n tin g s tra teg ie s a t stud y and te s t in fa c ial re co gn it io n .
A m erica n Jou rna l o f P sy ch o log y , 93 , 1 07 ±1 1 7 .
O ’ T o o le , A .J ., M il lw a rd , R .B ., & A nd e rson , J .A . (1 98 8). A p hy s ica l sys tem ap p roach to re co g n i tio n
m em ory fo r sp a tia lly t ra nsfo rm ed fa ce s . N eu ra l N etw o rk s , 1 , 1 79 ±1 9 9 .
Pac teau , C ., & B on th o ux , F . (19 94 ). T ra item en ts an a ly tiq u e s e t h o lis t iqu e s d e s v isage s [A na ly t ic
and ho lis tic p roce sse s o f fa c es ]. P sych o log ie F ra ncË a i se , 39 , 2 75 ±2 8 5 .
Pa rk in , A .J ., & G oo dw in , E . (19 8 3). T h e in flu ence o f d iffe re n t p ro ce ss in g st ra teg ie s on the
re cog n i tio n o f tra n sfo rm ed an d u ntr an sfo rm ed fa c e s . C an a dian Jo urn a l o f P sych o log y , 37 , 27 2 ±
27 7 .
Pa rk in , A .J ., & H ay w a rd , C . (1 98 3). T h e in fluen ce of t ra it a nd p h ys ica l- fea tu re -ba sed o rie n tin g
s tra teg ie s on a sp ec ts of fa c ia l m em o ry . B rit ish Jo urn a l o f P sych o logy , 74 , 7 1±8 2 .
Pa t te rson , K .E ., & B ad de ley , A .D . (19 77 ). W h en fa c e re cog n i tio n fa i ls . Jou rna l o f E xp e r im enta l
P sy cho lo g y: H u m a n Learn in g a nd M em o ry , 3 , 4 0 6±41 7 .
Pen ry , J. (1 9 71 ). L oo k in g a t fac e s an d rem em ber in g them : A gu id e to fac ia l iden ti fica tio n . L o nd on :
E lek B o ok s .
P igo tt , M .A ., & B rig h am , J.C . ( 19 85 ). T h e r el a tio n sh ip be tw een ac cu ra cy o f prio r d e scr ip t io n an d
fa c ia l re co gn ition . Jou rna l o f A p p lied P sycho lo gy , 7 0 , 54 7±5 55 .
P ro u dfo o t , R .E . ( 19 82 ). H em isp h eric a sy m m et ry fo r fa c e re co g n i tio n : S om e e ffe c ts o f v isu a l
m ask in g , hem ire t in al st im u la tion an d le arn in g ta sk . N eu ro p sych o lo g ia , 20 , 1 29 ±14 4 .
R ey no ld s , J .K ., & Pezd ek , K . (1 9 92 ). F ace re co g n i tio n m em o ry : T h e e ffe c ts o f ex po sure d ur at io n
and enco d ing in st ruc t io n . A pp lied C o gn itiv e P sych o log y , 6 , 2 7 9±29 2 .
Sch oo le r, J .W ., & E ng s tler -S ch oo le r, T .Y . (19 90 ). V erb a l ov e rshad ow in g o f v isua l m em o rie s : S om e
th in gs a re be tte r le ft u n sa id . C o g nit iv e P sycho lo gy , 2 2 , 36 ±7 1 .
Se rg en t , J. (1 98 6 ). M ic ro gen e sis o f fa c e pe rc ep tio n . In H .D . E ll is , M .A . Je ev es , J . N ew co m be , &
A .W . Y o un g (E d s .), A sp ec ts o f fa c e p ro ce ssing (p p .17 ±33). D o rd re ch t, T h e N e th erl and s : N ijho ff.
S e rv os , P ., & Pe ter s, M . (1 99 0 ). A cl e ar le ft h em isph e re adv an tage fo r v isu o -sp a tial ly ba sed ve rb a l
c a tego riz a tion . N eu ro p sycho lo g ia , 2 8 , 12 5 1±12 6 0 .
Sh ap i ro , P .N ., & Pen rod , S . (1 9 86 ). M e ta -a na lys is o f fa c ia l id en tific at io n s tu d ie s. P sycho lo g ica l
B u llet in , 10 0 , 13 9 ±1 56 .
Sh eph e rd , J .W . (19 8 9). T h e fa c e an d so ci a l a t trib u tio n . In A .W . Y ou n g & H .D . E l lis (E d s .),
H an d bo ok o f re sea rch on fa ce pro ce ss in g (p p .28 9±3 20 ). A m ste rdam : E lsev ie r.
S im pso n , W .E ., & C ra n da ll, S .J . (1 97 2) . T he pe rc ep tio n o f sm ile s . P sych on o m ic S c ience , 2 9 , 19 7 ±
20 0 .
Sm ith , A .D ., & W ino gra d , E . (1 97 8) . A d ul t ag e d i ffe re nce s in re m em be ring fa c e s . D evelop m en ta l
P sy cho lo g y , 1 4 , 4 43 ±4 4 4 .
Sp o rer , S .L . (1 9 91 ). D eep -D eep er- D eep e st? E n cod in g s tra teg ie s an d the re cog n i tio n o f hu m an
fa c e s. Jo u rn a l o f E xp er im en ta l P sy cho lo gy : Lea rn ing , M em o ry a nd C o gn it io n , 1 7 , 32 3±3 33 .
S t rin g e r, M . (1 9 8 7 ) . R eco g n i t io n m em o ry fo r fa c e s fo l lo w in g d e n o m in a tio n a l ju d g m en ts .
P ercep tua l a nd M o to r S k il ls, 65 , 2 76 .
S t rnad , B .N ., & M u e lle r, J.H . (1 97 7 ). L ev e ls o f pro ce ss ing in fa ci a l re cog n ition m em o ry . B ul le tin o f
th e P sycho no m ic So c ie ty , 9 , 1 7±1 8 .
S t roo p , J .R . (1 93 5 ). S tu d ies o f in te rfe re nce in se ri a l v e rba l re a c tio ns . Jou rna l o f E xp e r im enta l
P sy cho lo g y , 1 8 , 6 43 ±6 6 2 .
FACE RECOGNITION 567
T e rry , R .L . (1 9 93 ). H o w w ea rin g ey eg la sse s a ffe c ts fa c ia l re co g n i tio n . C u rren t P sych o logy :
D eve lo pm en ta l, L earn in g , P e rso na li ty, So c ia l, 1 2 , 15 1 ±1 62 .
T ib e rgh ien , G . (1 98 3 ). L a m eÂm o ire de s v isag e s [T he m em ory of fa c e s]. L ’ A n n eÂe P sych o log iqu e , 8 3 ,
1 53 ±19 8 .
T ib e rgh ien , G . (19 86 ). C on tex t e ffe c ts in re co g n i tio n m em o ry of fa ce s : Som e theo re tica l p ro b lem s.
In H .D . E llis , M . Jee v e s , F . N ew co m be , & A .W . Y o u ng (E d s .), A spec ts o f fa c e pro ce ss in g
(p p .88 ±104 ). D o rd r ec h t, T h e N e th er land s : N ijh o ff.
V a len tine , T ., & B ru ce , V . (1 98 6) . T he e ffe ct o f ra c e, inv e rs io n , and en cod in g ac tiv ity up on fa c e
re co gn ition . A c ta P sych o log ica , 6 1 , 25 9 ±2 7 3 .
V an K leeck , M .H . (1 9 89 ). H em isp h eric d i ffe re nces in g lo b a l v e rsu s loc a l p ro ces s in g o f h ie ra rc h ic al
v isu a l s t im ul i by n orm a l su b je c ts : N ew da ta and a m e ta -a n alys is o f p re v iou s s tu d ie s .
N eu rop syc ho lo g ia , 27 , 1 16 5±1 17 8 .
W a rrin g ton , E .K ., & A ckroy d , C . (19 7 5). T h e e ffe c t o f ori en ting ta sk s on r ec o g n it ion m em o ry .
M em ory & C og ni tio n , 3 , 14 0 ±1 42 .
W e lls , G .L ., & H ry c iw , B . (1 98 4). M em o ry fo r fa c e s : E n cod in g an d re t rie va l o p era t io ns . M em o ry &
C o g nit io n , 1 2 , 33 8 ±3 44 .
W e lls , G .L ., & T u rtl e , J .W . (19 88 ). W h a t is the be s t w ay to enco d e fa c es ? In M .M . G ru n eb e rg , P .E .
M orri s, & R .N . S yk e s (E ds .) , P ra c tica l asp ec ts o f m em o ry : C urren t re search a n d issu e s: V o l . I.
M em ory in e v e ryda y life (p p .1 63 ±1 6 8). C h ich e s ter, U K : W iley .
W in og ra d , E . (1 97 6 ). R ec o gn it io n m em o ry fo r fa c e s fo llow in g n ine d i ffe re n t ju dg m ents . B ul le tin o f
the P sycho n om ic S oc ie ty , 8 , 4 1 9±42 1 .
W in og ra d , E . (1 97 8) . E nco d ing op e ra tio ns w h ich fa c ili ta te m em ory fo r fa c e s a c ro ss th e life sp an . In
M .M . G run ebe rg , P .E . M o rris , & R .N . Sy k e s (E d s.), P ra c tica l asp ect s o f m em ory (p p . 2 55 ±2 6 2).
N ew Y ork : A cadem ic P re ss .
W in og ra d , E . (1 98 1 ). E lab ora tion an d d is tin c tiven e ss in m em ory fo r fa ce s . Jou rna l o f E xpe r im en ta l
P sych o logy : H u m an Learn in g a nd M em o ry , 7 , 18 1 ±1 90 .
W oo d head , M .M ., B ad de ley , A .D ., & S im m o nd s , D .C .V . (1 9 79 ). O n tra in in g p eo p le to re cog n ize
fa c e s . E rg on om ics , 2 2 , 33 3±3 43 .
Y o un g , A .W ., & E llis , H .D . (1 98 9) . S em an tic p ro ce ss in g . In A .W . Y o u ng & H .D . E ll is (E ds .),
H a nd bo o k o f re sea rch o n face p ro ce ssing ( pp .2 35 ±2 62 ). A m ste rd am : E lsev ie r.
568 COIN AND TIBERGHIEN