20.11.2012 Views

Winkielman, Zajonc, & Schwarz, 1997

Winkielman, Zajonc, & Schwarz, 1997

Winkielman, Zajonc, & Schwarz, 1997

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

C O GNIT IO N A ND E M OT IO N, 1 9 9 7 , 1 1 (4 ), 4 33 ± 4 6 5<br />

S ublim inal A ffective P rim ing R esists<br />

A ttribution al Interven tions<br />

P io tr W in kielm an<br />

U n iversity o f M ich ig a n, A nn A rb o r, U S A<br />

R o bertB . Z ajo n c<br />

S ta n fo rd U n ive rsity , U S A<br />

N o rbertS chw arz<br />

U n iversity o f M ich ig a n, A nn A rb o r, U S A<br />

W e exam ine tw o explanati ons of the subl im inal affe ctiv e prim ing effect. The<br />

fe eling s-as-inf orm ation m ode l (S chw arz & Clore , 1 988) holds that judg em<br />

ents are bas ed on pe rcepti ble fe e ling s. Henc e, affe ctive in¯ uences de pe nd<br />

on the sourc e to w hich fe eling s are (m is)attribute d. In contras t, the affe ctiv e<br />

prim ary hy pothe sis (<strong>Zajonc</strong> , 1 980) sug g ests that affe ctiv e in¯ ue nc es should<br />

resis t attributional inte rve nti ons. This is be caus e the affective sy stem responsible<br />

for pre fe renc es is separate from the cog nitiv e syste m respons ibl e for<br />

inf erenc e s; be caus e early affe ctiv e proc esses are autom atic and the refore<br />

inac cessibl e to hig he r-orde r inte rventi ons; and be cause early affe ctiv e<br />

respons es are not re pre sente d as cons cious fe eling s. W e te ste d the se explanati<br />

ons in tw o experim ents that crossed subl im inal affe c tive prim ing w ith<br />

(m is)attribution m anip ulati ons . B oth studi es found reliabl e shif ts in judg em<br />

ents of ne utral stim uli as a result of prim es even w he n subj ects w ere aw are<br />

that the ir feeling s m ig ht not be diag nostic for the judg em ent at hand.<br />

S ubje cts did not report ex pe rienc ing any fe eling s in response to the<br />

prim es. The obtai ne d affe ctiv e prim ing effe ct w as inde pe nde nt of response<br />

tim es and subje ctiv e reports of eng ag ing in judg em ental corrections . How -<br />

ever, the prim ing effect did prov e sensitive to the expe rim ental instruc tion s.<br />

W e discus s the im pli cations of the se ® ndi ng s for the affe ctiv e prim ac y<br />

hy pothe sis and the fe eling s-as-inf orm ation m ode l.<br />

Re que sts for re p rin ts should be se nt to Piotr W inkie lm an, D epartme nt of Psy c holog y ,<br />

Univ e rsity of M ic hig an, A nn A rbor, MI 4 8 1 0 9 ± 1 1 0 9 , US A ; e -m ail m ay be se nt to<br />

w in kpio @ um ic h.e du.<br />

The au thors w ou ld li ke to thank K ent B e rri dg e , Dov C ohe n, Mare k Drog osz , Phoe be<br />

Ells w orth, Ke nt Harbe r, B arbe l Knaupe r, M e lv in Manis, Danny M c Intosh, S he il a M urphy ,<br />

C hris Ne ls on, A ndrz e j Now ak, A le x Rothm an, K evin Tay lor, an d Phil W ong f or their he lp at<br />

v arious stag e s of this proje c t. In partic ular, w e thank Jerry C lore and se v e ral anony m ous<br />

re v ie w e rs for the ir v e ry he lpful c om m e nts on a pre v io us v e rsion of this artic le .<br />

Ó 1 9 9 7 Psy c holog y Pre ss, an im pri nt of E rlb aum (U K) Tay lor & Franc is Ltd


434 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

INTR ODUCTIO N<br />

Num e rous studie s hav e dem ons trate d a pronounc e d im pac t of affe c tiv e<br />

state s on e v aluati v e judg e m e nts (for rev ie w s se e C lore , S c hw arz , & Conw<br />

ay , 1 9 9 4 ; Forgas, 1 9 9 5 ; S c hw arz & Clore, 1 9 9 6 ). In the prese nt pape r, w e<br />

addres s a spe ci® c aspe c t of the interplay be tw ee n affe c t and c og nition,<br />

nam e ly , how e x pos ure to affe c tiv e stim uli presente d be low the threshold<br />

of aw are ne ss in¯ ue nc e s e v aluations of unrelate d targ e t s tim uli. In a rec e nt<br />

study , Murphy and <strong>Zajonc</strong> (1 9 9 3 ) e x pose d subje c ts to supralim inally prese<br />

nte d ne utral stim uli in the form of C hine se ide og raphs and aske d the m to<br />

m ake a liking judg e m e nt (study 1 ) or to judg e w he the r the ide og raphs<br />

represe nte d ``g ood’ ’ or ``bad’ ’ obje c ts (s tudy 2 ). T he subje c ts did not<br />

know that the ideog raphs w e re prece de d by affec tiv e and nonaffe c tiv e<br />

sublim inal prim es. A ffe c tiv e prim e s w e re pic tures of s m iling or ang ry<br />

fac e s, w he reas nonaffe c tiv e prim e s w e re pictures of random poly g ons.<br />

S ublim inal affe c tiv e prim es produc e d sy stem atic shifts in s ubje c ts ’ judg e -<br />

m e nts of the ide og raphsÐ ide og raphs prec e de d by sm iling fac e s w ere like d<br />

the m ost and ide og raphs prec e de d by frow ning fac e s w e re like d the le as t,<br />

w ith ide og raphs prec e de d by poly g ons falling in be tw e e n. Murphy , Monahan,<br />

and <strong>Zajonc</strong> (1 9 9 5 ) replic ate d the s e e ffe c ts in a m ore e x te nsiv e study .<br />

De spite num e rous studie s that hav e found affe c tiv e prim ing , the m e diating<br />

proc e ss has rec e iv e d little atte ntion. Previous researc h foc use d prim arily<br />

on de m onstrating that affe c tiv e prim ing c an be ac hie v e d w ith m inim al<br />

stim ulus e x pos ures (e .g . B arg h & Pie trom onac o, 1 9 8 2 ; Murphy & <strong>Zajonc</strong> ,<br />

1 9 9 3 ; Nie de nthal, 1 9 9 0 ), and on e x ploring the inte rac tions be tw e e n c og -<br />

nitiv e and affe c tiv e proc e sse s (e .g . Edw ards, 1 9 9 0 ; Kitay am a, 1 9 9 1 ). The<br />

present pape r addresse s the affe c t-judg e m e nt link m ore direc tly . W e s ee k<br />

to de te rm ine w he the r sublim inal affe c tiv e prim ing effe c ts are m ediate d by<br />

c ontrollable proc e ss es inv olv ing infe renc e s from consc iously e x pe rie nce d<br />

affe c t, or by autom atic proc e sse s im pe rvious to hig he r-orde r inte rve ntions.<br />

W e be g in by c om paring tw o e x planati ons of sublim inal affe c tiv e prim ing<br />

e ffe c ts and subse que ntly report tw o studie s de sig ne d to te st the se c ontrasting<br />

e x planations.<br />

A Feelings-as-inform ation Account of<br />

Affective Prim ing<br />

A c c ording to the fe e ling s-as -inform ation m ode l, affe c tiv e in¯ ue nc e s oc cur<br />

be c aus e judg e m ents are base d on pe rceptible affe c tiv e state s. S c hw arz and<br />

C lore (1 9 8 3 , 1 9 8 8 , 1 9 9 6 ; Clore, 1 9 9 2 ) propose d that pe ople m ay s im plify<br />

c om ple x judg e m ental tas ks by turning to the ir appare nt affe ctiv e reac tion<br />

to the targ e t, e ss e ntially as king the m se lv e s, ``How do I fe e l about it? ’ ’ In so<br />

doing , the y m ay m isread pre-e x is ting fe e ling s as a reac tion to the targ e t,


SU B LIM IN A L A FFECT IV E P RIM IN G 435<br />

resulting in m ore positiv e ev aluations w he n in a positiv e rathe r than<br />

ne g ativ e affe c tiv e s tate . Referring to an e arly v ersion of the Murphy and<br />

<strong>Zajonc</strong> pape r (1 9 9 3 ), S c hw arz (1 9 9 0 , p. 5 3 8 ) spe c ulate d on how the<br />

fe e ling s -as-inform ation m ode l m ay e x plain s ublim inal affe c tiv e prim ing<br />

e ffe c ts: ``Giv e n the abs enc e of any use ful know le dg e about the ide og raph,<br />

subje c ts m ay be like ly to turn to the ir affe c tiv e response , asking the m -<br />

se lv e s, `How do I fe e l about it? ’ If the y e ncounte r positiv e fe e ling s , the y<br />

m ay c onc lude that the ide og raph m ay m ean s om e thing pos itiv e Ð unle ss<br />

the y hav e reason to doubt the inform ati onal v alue of the ir fe e ling s.’’<br />

Ac c ording to the fe e ling s-as-inform ation m odel, judg e m e nts are only<br />

base d on aff ec tiv e e x pe rie nc e s if the e x pe rie nc e s are pe rce iv e d as diag -<br />

nostic for the judg e m e nt at hand. If the inform ational v alue of the e x pe rie<br />

nc e is c alle d into que stion, for e x am ple , if s ubje c ts (m is)attribute the ir<br />

fe e ling s to a s ource unre late d to the targ e t, the im pac t of affe c tiv e state s on<br />

judg e m ent is e lim inate d (e .g . S c hw arz & C lore, 1 9 8 3 ). This m ode l assum e s<br />

relianc e on one ’ s fe eling s as a d efa u lt that doe s not require c onsc ious<br />

attribution of the fe e ling to the obje c t of judg e m e nt, but holds that<br />

c onsc ious attribution of the fe e ling to an irrele v ant source e lim inate s the<br />

othe rw ise obs erv e d affe c tiv e in¯ uenc e .<br />

Althoug h m any s tudie s support the predic tions of the fee ling -as-inform<br />

ati on m ode l for the im pac t of m oods and othe r phenom e nal e x pe rie nc e s<br />

(se e Clore e t al. 1 9 9 4 ; S c hw arz & C lore, 1 9 9 6 for revie w s ), it is an ope n<br />

issue w hethe r the m ode l applie s to sublim inal prim ing e ffe c ts. The re are<br />

g ood reasons to think it m ig ht. C learly , the proc e dures use d in sublim inal<br />

prim ing ex pe rim ents g uarantee that subje c ts are not aw are of the s ource of<br />

the ir affe c tiv e reac tion. He nc e , the y are like ly to pe rce iv e the ir ow n affe c t<br />

as a response to the supralim inally prese nte d targ e t, thus rende ring it<br />

rele v ant to the judg e m e nt. If so, m aking subje c ts aw are of the pote ntial<br />

im pac t of s ublim inal prim e s should e lim inate the im pac t of affe c tiv e<br />

prim ing , be c ause it rende rs salie nt an irrele v ant source of their affe c tiv e<br />

e x pe rie nc e . Consis te nt w ith this predic tion, Murphy and <strong>Zajonc</strong> (1 9 9 3 )<br />

obtai ne d no prim ing e ffe c t w he n the y prese nte d sm iling or frow ning fac e s<br />

supralim inally , sug g e s ting that aw are ne ss of the source of one’ s fe e ling s<br />

e lim inate d the ir im pac t. He nce , the av ailable ® nding s are c om patible w ith<br />

a fe e ling s-as-inform ati on ac count of affec tiv e prim ing e ffe c ts, althoug h a<br />

spe c i® c e m piric al te s t is m iss ing .<br />

An Affective Prim acy Account of Affective Prim ing<br />

In 1 9 8 0 , <strong>Zajonc</strong> propos e d the affe c tiv e prim ac y hy pothe sis (<strong>Zajonc</strong> , 1 9 8 0 ,<br />

1 9 8 4 , 1 9 9 4 ). The hy pothe sis holds that affe c tiv e reactions require m inim al<br />

stim ulus e laboration and c an oc c ur v e ry quic kly . T he hy pothe sis also<br />

sug g e sts that the affe c tiv e sy ste m , c onc e rne d w ith s tim ulus e v aluation, is


436 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

se parate and partially inde pe nde nt from the c og nitiv e sy ste m c onc e rne d<br />

w ith stim ulus m eaning . Translati ng <strong>Zajonc</strong> ’ s hy pothe sis into the lang uag e<br />

propose d by Fodor (1 9 8 3 ), the affe c tiv e s y ste m represe nts a m e ntal m odule<br />

. S uc h a m odule is dom ain-spe ci® c (c onc e rne d w ith the e v aluation of<br />

e m otional sig ni® c anc e ), ope rate s on pre-se m antic input (be fore the le v e l of<br />

m e aning ), encapsulate d (inde pe nde nt from m odule s that proc e ss othe r<br />

attribute s), and c og nitiv e ly im pe ne trable (im perv ious to hig he r-orde r influe<br />

nc e s). The assum ptions of the affe c tiv e prim ac y hy pothe s is are c onsiste nt<br />

w ith psy c holog ic al and ne urophy s iolog ic al e v ide nc e (Ag g le ton & Mishkin,<br />

1 9 8 6 ; Dam asio, 1 9 9 4 ; Kunst-W ilson & <strong>Zajonc</strong> , 1 9 8 0 ; L eDoux , 1 9 8 9 ,<br />

1 9 9 3 ).<br />

The affe c tiv e prim ac y hy pothe sis predic ts inde pe nde nc e of sublim inal<br />

affe c tiv e prim ing from attributional m anipulations for s e v e ral reasons.<br />

A ttributional inte rve ntions atte m pt to c hang e prefe renc e s by targ e ting<br />

c ons c ious infe renc es . Inferenc es , how e v e r, are a part of the hig h-orde r,<br />

se m antic s y ste m that has no direc t ac c e ss to the low -order, nonse m antic<br />

prefe renc e s y ste m . C ons iste nt w ith this assum ption, studie s on autom ati c<br />

v ig ilanc e (Pratto, 1 9 9 4 ; Pratto & John, 1 9 9 1 ) and autom atic e v aluation<br />

(B arg h, C haike n, Gov e nde r, & Pratto, 1 9 9 2 ) found that the ope ration of<br />

affe c tiv e proc e ss e s is hig hly inde pe nde nt of infe rential g oals (B arg h,<br />

1 9 8 9 ).<br />

Note , how e v er, that sublim inal affe ctiv e prim ing should n o t be e x pe c te d<br />

to be inde pe nde nt of all aspe c ts of attributional m anipulati ons. Thes e<br />

m anipulations should hav e no in fe re n tia l e ffe c ts , y e t the y m ay in¯ ue nc e<br />

suc h v ariables as atte ntion or affe c tiv e state s that, in turn, m ay m odify<br />

affe c tiv e prim ing v ia affe c t-affe c t inte rac tions (M urphy , Monahan, &<br />

<strong>Zajonc</strong> , 1 9 9 5 ) or atte ntion-affe c t inte rac tions (Lang , 1 9 9 5 ).<br />

A nothe r reason w hy , ac c ording to the affe c tiv e prim ac y hy pothe sis,<br />

attributional inte rve ntions should be ine ffe c tiv e is relate d to the ass um ptions<br />

about the form of early affe c tiv e e x pe rie nc es . A ny suc ce ss ful m anipulation<br />

of the infe renc e s that subje c ts m ay draw from their fe e ling s requires<br />

som e aw are ne ss of thes e fe e ling s in the ® rst plac e . How e v e r, affe c tiv e<br />

prim ing m ay n o t inv olv e a c hang e in c onsc ious ly e x pe rie nc e d fe e ling s.<br />

First, som e affe c tiv e in¯ ue nc e s on pe rce ption, inte rpretation, and de c is ion<br />

m ay be m e diate d by low -le v e l sy ste m s that do not produc e any ac c om pany<br />

ing consc ious affe c t at all (Dam as io, 1 9 9 4 ). S ec ond, e v e n if affe c tiv e<br />

prim ing le ads to a notice able affe c tiv e reac tion, this reaction m ay not be<br />

represe nte d as a fe e ling , but only as a chang e in a prefe renc e . R ecent<br />

the orising sug g e sts that the sy ste m that e v aluate s the e m otional sig nific<br />

anc e of stim uli is faste r than the s y ste m that elic its fe e ling s (Le Doux ,<br />

1 9 9 3 ). Ladav as, Cim atti, DelPe sc e , and Tuoz z i (1 9 9 3 ) found that e v aluativ<br />

e disc rim ination response s (presum ably unde rly ing prefe renc e s) w e re<br />

obtaine d w he n s ublim inal affe c tiv e s lide s w e re prese nte d to the rig ht and to


the le ft he m isphe re of a split-brain patie nt. How e v e r, autonom ic response s<br />

(presum ably unde rly ing fe e ling s) w e re only obtaine d w he n sublim inal<br />

slide s w e re prese nte d to the rig ht hem isphe re. T he le ft he m isphe re, w hic h<br />

is le s s able to respond to e m otional stim uli, produc e d autonom ic response s<br />

only for slide s prese nte d supralim inally .<br />

In sum , the m inim al e x posure c onditions of sublim inal prim ing m ay<br />

e lic it w e ak, undiffe rentiate d prefe renc e response s, but m ay not s upport the<br />

am ount of proc e s sing ne c e ssary to e lic it identi® able fe e ling s. More g e ne<br />

rally , prefe renc e s se e m to represe nt w hat Fodor (1 9 8 3 ) c alls ``s hallow<br />

output’ ’ (i.e. the y are prim itiv e , pre-se m antic , and unappraise dÐ uninte rprete<br />

d). A s <strong>Zajonc</strong> (1 9 9 4 ) sug g e ste d, an e arly , unappraise d affe c t m us t be<br />

e laborate d and inte g rate d w ith the output of othe r m odule s to result in an<br />

ide nti® able fe e ling . In that res pe c t, s ublim inal affe c tiv e prim ing m ay be<br />

sim ilar to the m e re e x posure paradig m , w he re s ubje c ts do not pe rce iv e<br />

the m se lv e s as c hang ing prefe renc es by v irtue of e x posure.<br />

The prec e ding disc ussion indic ate s that sublim inal prim ing m ay not<br />

e lic it fe eling s that c an be use d as input into an infe rential strate g y . S e v e ral<br />

approac hes to the affec t-judg em e nt link share this assum ption but diffe r<br />

regarding the nature and loc us of the proc e sse s that m e diate be tw e e n affe c t<br />

and judg e m e nt. S om e ac c ounts are c ons iste nt w ith the affec tiv e prim ac y<br />

hy pothes is and postulate low -orde r, s ensory, nons em antic proc e sse s. Othe r<br />

ac c ounts assum e hig h-orde r se m antic m ediation. Moreov e r, the ac c ounts<br />

loc ate the c ritic al proc e ss at diffe rent s tag es Ð atte ntion, pe rce ption, inte rpretati<br />

on, or g e ne ration of m otor response . First, rese arc h into the affe ctatte<br />

ntion link propose s that affe c tiv e prim es lead to he ig hte ne d atte ntion to<br />

affe c t-c ong rue nt inform ation (se e De rrybe rry & T uc ke r, 1 9 9 4 for a<br />

revie w ). S e c ond, rese arc h into the affe c t-pe rce ption link s ug g e sts that<br />

affe c tiv e prim e s trig g e r ac tiv ation in pe rce ptual m e m ory, w hic h results<br />

in a m ore e f® c ient pe rce ption of affe c t-c ong rue nt targ e ts (se e Nie -<br />

de nthal, S e tte rlund, & Jone s, 1 9 9 4 for a rev ie w ). Third, s e m antic netw ork<br />

m ode ls sug g e st that affe c tiv e prim e s bias inte rpretati on of targ e ts by<br />

ac tiv ati ng v ale nc e -c ong rue nt c onc e pts (se e B ow e r, 1 9 9 1 for a rev ie w ).<br />

Finally , affe c t e licite d by prim e s m ay m odify judg e m e nt be hav iour v ia<br />

c onne c tions be tw e e n the affe c tiv e sy stem and the m otor response sy ste m<br />

(se e <strong>Zajonc</strong> & Markus, 1 9 8 4 for a revie w ). W e return to the se issue s in the<br />

Disc ussion se c tion.<br />

Overview of Studies<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 437<br />

As the e arlier disc us sion illustrate s , se v e ral line s of e v ide nc e are com patible<br />

w ith the notion that affe c tiv e in¯ ue nc e s on judg e m ent and be hav iour<br />

m ay be m e diate d by m e c hanism s othe r than the m isattribution of e x pe rie<br />

nc e d fe e ling s. How e v e r, until the role of the se m e c hanism s in sublim inal


438 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

prim ing is m ore prec ise ly s pe c i® e d, sy ste m ati c e m piric al tes ts of the s e<br />

m e c hanism s rem ain dif® c ult. In c ontrast, the fe e ling s-as-inform ation<br />

m ode l g e ne rate s predic tions that are am enable to e m piric al te sting .<br />

He nc e , the prese nt studie s foc us on a te st of the predic tions g enerate d by<br />

the fe e ling s -as-inform ation acc ount. S pe ci® c ally , w e inform ed subje c ts in<br />

som e c onditions of our e x pe rim ents that the y w ould be e x pos ed to sublim<br />

inal prim es that m ig ht elic it a fe e ling . De pe nding on c onditions, subje<br />

c ts w e re e ither inform e d that this reac tion w ould be positiv e , inform e d<br />

that it w ould be ne g ative, or w e re not inform e d about its v alence . Thes e<br />

attribution m anipulati ons w e re introduc e d as a be tw e e n-subjec ts fac tor and<br />

w e re c ross e d w ith the v ale nc e of the sublim inal prim es, w hic h w as<br />

m anipulate d w ithin subje c ts.<br />

B as e d on the se m anipulations, the fe eling s -as -inform ation m ode l predic<br />

ts the e m e rge nc e of augm e ntation and dis c ounting e ffe c ts (Ke lle y ,<br />

1 9 7 2 ). W he n subje c ts ex pe c t sublim inally prese nte d prim e s to e lic it a<br />

positiv e fe e ling , the y should (correc tly ) attribute any pos itiv e fe e ling<br />

the y e x pe rie nc e to the im pac t of the prim ing m anipulation (i.e . to a<br />

sublim inally prese nte d sm iling face ). S im ilarly , w he n subje c ts e x pe c t<br />

sublim inally presente d prim e s to elic it a ne g ative fe e ling , the y should<br />

(c orrec tly ) attribute any ne g ativ e fe e ling the y e x pe rie nc e to the im pac t<br />

of the prim ing m anipulation (i.e. to a sublim inally prese nte d frow ning<br />

fac e ). Giv e n that s ubje c ts are told that the prim ing m anipulation is unrelate<br />

d to the judg e m ent tas k, the y should disc ount the inform ational v alue of<br />

the ir on-line affe c tiv e e x pe rie nc e and thus not e x hibit the sublim inal<br />

prim ing e ffe c t. In contrast, if an affe c tiv e e x pe rie nc e e licite d by the<br />

sublim inal prim es is inc onsistent w ith subje c ts’ e x pe c tati ons, the e x pe rie<br />

nc e c annot be attribute d to the e x pe rim ental m anipulation and should<br />

se e m partic ularly diag nostic , resulting in inc rease d s ublim inal prim ing<br />

e ffe c ts. A c cording ly , a dis counting e ffe ct is predic te d w he n subje c ts ’<br />

e x pe rie nc e s m atc h the ir e x pe c tations and an aug m entati on e ffec t is predic<br />

te d w he n their e x pe rie nc e s c ontradic t their ex pe c tati ons. Finally , w he n<br />

subjec ts are inform e d that s ublim inally prese nted m ate rials m ay e lic it an<br />

unspe c i® e d reaction, the y m ay doubt the inform ati onal v alue of any reac -<br />

tion the y m ay e x pe rienc e , thus reducing sublim inal prim ing e ffe c ts inde -<br />

pe nde nt of the v ale nc e of the ir e x perie nc e .<br />

Note that the aforem e ntione d predic tions of dis counting a n d aug m e ntation<br />

e ffe c ts are inc om patible w ith an alte rnative se t of predic tions that m ay<br />

be de riv e d from the im plic ations of the e x pe rim ental ins truc tions, inde pende<br />

nt of subje c ts ’ af fe c tiv e e x pe rie nc e s. A s S track and c olleag ue s (S trac k,<br />

1 9 9 2 ; S trac k, S c hw arz , B le ss, KuÈ ble r, & W aÈ nke , 1 9 9 3 ; se e also W e g e ne r &<br />

Pe tty , 1 9 9 5 ) de m onstrate d, subje c ts m ay c orrec t judg e m ents to c om pe nsate<br />

for in¯ ue nc e s sug g e ste d by the e x pe rim e nter. If so, subje c ts w ho are<br />

inform e d about positiv e in¯ uenc es should correc t all judg e m e nts dow n-


w ard, w he reas s ubje c ts w ho are inform e d about ne g ative in¯ ue nc e s should<br />

c orrec t all judg e m e nts upw ard, inde pe nde nt of the ir actual af fe c tiv e e x pe rie<br />

nc e . In c ontrast to suc h a m ain e ffe c t of e x pe c tations, the fe e ling s-asinform<br />

ati on m ode l predic ts an inte rac tion of e x pe c tati ons w ith ac tual<br />

affe c tiv e ex pe rienc e s, as de sc ribe d e arlie r.<br />

The affe c tiv e prim ac y hy pothe sis predic ts that the attributional m anipulations<br />

should not produc e the aforem e ntione d patte rn of judg e m e nts.<br />

This prediction re sts on three assum ptions. First, the sy ste m inv olv e d in<br />

affe c tiv e prim ing is resistant to infe renc e s, be cause infe renc es are se m antic<br />

w he reas prefe renc e s are nons e m antic. S ec ond, affe ctiv e prim ing is autom<br />

ati c and should be e nc apsulate d from any hig he r-orde r inte rve ntion.<br />

Third, affe c tiv e c hang e s m e diating affe c tiv e prim ing are e ithe r not represe<br />

nte d c ons c iously at all, or at le ast the y are not repre se nted as fe e ling s.<br />

He nc e , subje c ts are not aw are of a fe e ling e x pe rienc e that m ay be disc<br />

ounte d or aug m e nte d as a source of inform ati on in form ing a judg e m ent<br />

about the targ e t.<br />

Giv e n that the affe c tiv e prim ac y hy pothe sis predicts a null e ffe ct of<br />

attributional inte rve ntions , w e did not w ant to reje c t the fe e ling s-as-inform<br />

ati on m ode l bec ause of pote ntially w eak m anipulations, resulting in the<br />

ac c e ptan c e of the affe c tiv e prim ac y hy pothe sis as a de fault. The refore, w e<br />

took the follow ing s te ps : Our attributional m anipulations w e re c le ar and<br />

e x plicit; the attributional m anipulati on w as chang e d from s tudy 1 to study<br />

2 ; and both s tudie s independe ntly te ste d for the prese nc e of affe c tiv e<br />

e x pe rie nc e s. The robus tnes s of affe c tiv e prim ing w ill be de m ons trate d if<br />

it is obtained e v e n w he n the e x pe rim ent offe rs s ubje cts a c le ar opportunity<br />

to disc ount the in¯ ue nc e of affe c tiv e prim e s.<br />

M ethod<br />

Subjects<br />

EXPER IM EN T 1<br />

S ix ty -three m ale and fe m ale underg raduate subje c ts at the Univ e rsity of<br />

Michig an participate d in partial ful® lm e nt of a c ourse requirem ent. S ubje cts<br />

w ho kne w Chines e , Korean, or Japanes e w ere e x c lude d from partic ipati on.<br />

M aterials and Apparatus<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 439<br />

Or d e r a n d typ e o f s tim u li. T he liking judg e m e nt tas k c onsiste d of the<br />

prese ntation of targ et ide og raphs prec e de d by v arious prim e s. The re w e re<br />

tw o bloc ks of 2 0 trials. Eac h bloc k e m ploy e d a diffe rent arrang em e nt of<br />

stim uli. In the ® rst bloc k of 2 0 trials the orde r of prim e s w ithin a bloc k w as


440 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

random ise d and c ons iste d of: (a) ® v e dis tinc t ide og raphs prec e de d by no<br />

prim es at all; (b) ® v e distinct ide og raphs prec e de d by pic tures of ® v e<br />

distinc t poly g ons; (c ) ® v e distinct ide og raphs prec e de d by prim e s show ing<br />

ang ry fac e s of ® v e distinc t indiv iduals; and (d) ® v e distinc t ide og raphs<br />

prec e ded by pic tures of happy fac e s of ® v e dis tinc t indiv iduals.<br />

In the se c ond bloc k, the prim es c onsiste d of: (a) ® v e distinc t ide og raphs<br />

prec e ded by no prim es at all; (b) ® v e distinc t ide og raphs prec e de d by<br />

pic tures of ® v e distinc t poly g ons; (c ) ® v e ide og raphs repe ate d from the<br />

® rst block w e re prec e de d by ang ry fac e s of the sam e indiv iduals that<br />

prov ide d happy e x pressions use d in bloc k 1 ; and (d) ® v e ide og raphs<br />

repe ate d from the ® rst bloc k w e re prec e de d by happy fac e s of the sam e<br />

indiv iduals that prov ide d ang ry e x press ions use d in bloc k 1 .<br />

Prim e s. Photog raphs of 1 0 indiv iduals se rve d as fac ial prim e s. E ac h<br />

indiv idual prov ide d tw o e x pressionsÐ happine ss and ang e r. The se 1 0<br />

indiv iduals w ere se le c te d from a se t of 6 5 slide s pre-te ste d to prov ide<br />

g ood e x am ple s of anger and happine s s.<br />

Stimu lu s p re se n ta tio n . The e x pos ure c onditions and the e quipm e nt w e re<br />

ide ntic al to the one us e d by Murphy and Z ajonc (1 9 9 3 ). Four slide<br />

proje c tors w ith a Uniblitz shutte r and a red ® lte r w ere use d to proje c t<br />

4 5 cm 3 6 0 c m im ag es on a sc ree n at subje c ts ’ e y e le v e l at a distanc e of<br />

approx im ate ly 1 .5 m . This presentati on resulted in a 1 7 -de g ree v is ual<br />

horiz ontal ang le and 2 0 -de g ree v e rtic al ang le . Lum inanc e of the s cree n<br />

® e ld w as approx im ate ly 6 0 c d/m 2 . The shutte rs w e re c ontrolle d by tw o<br />

Uniblitz relay control box e s c onne c te d to an XT -com pute r. The shutte r<br />

spe e d for the sublim inal prim e w as se t at 4 m se c w hic h, afte r adding ope n<br />

and shut de lay , results in a 1 0 m se c ¯ ash.<br />

Procedure and Design<br />

Se q u e n c e o f e v e n ts d u rin g a ffe c tiv e p r im in g . On arriv al at the laboratory,<br />

subjec ts w e re told that the study c onsiste d of tw o tas ks. In the ® rst tas k the y<br />

w ould m ake liking judg e m e nts of s e v e ral C hine se ideog raphs on a 6 -button<br />

response box . Num be r 1 on the box indic ate d that the subje c t did not like<br />

the ide og raph at all, w he reas num be r 6 indic ate d that it w as like d quite a<br />

bit. Eac h trial las te d for about 1 1 se c onds (se e Fig . 1 ). The sublim inal<br />

stim ulus (the fac e or the poly g on) w as prese nte d for 1 0 m se c and w as<br />

im m ediate ly follow e d by an ide og raph that also s e rve d as a backw ard<br />

m ask. T he S O A be tw e e n the prim e and the ide og raph w as approx im ate ly<br />

5 m se c . Eac h ide og raph stay e d on the sc ree n for 2 0 0 0 m se c . A fter an 8 -se c ond<br />

paus e for the subje ct’ s response , the ne x t pair c ons isting of sublim inal<br />

prim e and ideog raph w as show n. A fte r the ® rst tas k, the re w as a s hort<br />

break. T he se c ond tas k of the study te ste d w he the r the prim e s w e re<br />

presente d sublim inally (s ee late r).


SU B LIM IN A L A FFECT IV E P RIM IN G 441<br />

Ma n ip u la tio n o f in fe r en c e s. For the ® rst tas k of the s tudy , s ubje c ts w e re<br />

g iv e n one of the ® v e w ritte n instruc tions, de sig ne d to m anipulate be lie fs<br />

about the ir affe c tiv e reac tions during the e x pe rim e nt. In the ® rst c ondition,<br />

subje c ts w e re not inform ed about sublim inal s tim uli, but w e re sim ply aske d<br />

to m ake liking judg e m ents of ide og raphs. This c ondition w as de sig ne d as a<br />

sim ple replic ation of the Murphy and <strong>Zajonc</strong> (1 9 9 3 ) study 1 , and w e re fe r<br />

to this condition as a ``n o e xpe c ta tio n s’’ c ondition.<br />

In the ``n o n s p e c i® c e xpe c ta tio n c o n d itio n ’’ , instruc tions state d that<br />

shortly be fore the prese ntati on of ide og raphs othe r pic ture s w ould appe ar<br />

v e ry brie ¯ y on the sc reen. T he instruc tion m e ntione d that the se pic tures<br />

w ould appe ar s o brie ¯ y that no one w ould be able to se e the m c onsc<br />

iously . Possible affe c tiv e reac tions and the nature of sublim inal pic tures<br />

w e re not m e ntione d. To justify to subjec ts the prese nc e of sublim inal<br />

prim e s, the instruc tions refe rred to ``the se c ond part of the s tudy ’ ’ w he re<br />

the se slide s w ould be rele v ant. The purpose of the ``nonspec i® c e x pe c tation<br />

c ondition’’ w as to te s t w he the r the she e r fac t of be ing inform e d<br />

about poss ible prim e s w ould e lim inate the basic affe c tiv e prim ing e ffe c t.<br />

For this condition, and for all the rem aining c onditions , the affe c tiv e<br />

prim ac y ac c ount predic ts a replic ation of the prim ing e ffe c t. The fe e ling sas-inform<br />

ati on ac c ount predic ts that in the nonspe ci® c e x pe c tati on c ondition<br />

subje c ts m ay be c om e dis trus tful of the ir affe c tiv e reac tions , w hic h<br />

FIG . 1. S e q ue nc e o f e ve nts o n o ne tri al du rin g lik ing ju dge m en t tas k.


442 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

should unde rm ine the prim ing e ffe c t. Prim ing e ffe cts should be w e ake ne d<br />

to the e x te nt that s ubje cts thoug ht the se ``othe r pictures’ ’ m ig ht in¯ ue nc e<br />

the ir affe c tiv e state .<br />

In the ``p o sitiv e e xpe c ta tio n c o n d itio n ’’ subje cts w e re inform ed that<br />

sublim inal sm iling fac e s w ould brie ¯ y appe ar be fore s om e ide og raphs<br />

and m ay bring on ``s om e ple asant fe e ling s’ ’ and c ause s om e ``positiv e<br />

g ut reac tions’’ . For this c ondition the fe e ling -as-inform ati on ac c ount predic<br />

ts that subje c ts should disc ount the ir positiv e reac tions to the ide og raphs<br />

prec e ded by positiv e prim es , w hereas the ir judg e m e nts of ide og raphs<br />

prec e ded by ne g ative prim e s should c ontinue to be in¯ ue nc e d. Moreov e r,<br />

the judg e m e nts of ide og raphs prec e de d by ne g ati v e prim es should be<br />

e spe cially unfav ourable be c ause , by the log ic of the aug m e nting e ffe c t,<br />

subjec ts should tre at the ir ne g ativ e reac tions as e spe c ially diag nostic of<br />

the ir true liking response to the targ e t ide og raphs .<br />

In the `` n e g a tiv e e xp e c ta tio n c o n d itio n ’ ’ subje c ts w e re inform ed that<br />

sublim inal ang ry fac e s w ould appe ar be fore ide og raphs and m ay bring<br />

on ``som e unple asant fe e ling s’ ’ and c ause s om e ``ne g ativ e g ut reactions ’’ .<br />

Predic tions for this c ondition are the e x ac t reversal of predic tions for the<br />

positiv e ex pe c tati on c ondition.<br />

In the ``p o sitiv e a n d n e g a tiv e e xpe c ta tio n c o n d itio n ’ ’ subje c ts w e re<br />

inform e d that ang ry and happy fac e s w ould be show n and m ig ht c aus e<br />

c orresponding fe eling s and ``g ut re ac tions’ ’ . He re, assum ing suf® c ie nt<br />

disc ounting , the fe e ling -as-inform ation ac c ount predic ts a disappe aranc e<br />

of the prim ing e ffe c t, w he re as the affe c tiv e prim ary acc ount predic ts no<br />

suc h e ffe c t.<br />

F o r c e d c h o ic e . Follow ing 4 5 trials, subje c ts in all c onditions w e re g iv e n<br />

a force d-c hoic e te s t of aw are ne ss (s e e Fig . 2 for de tai ls ). S ubje c ts w e re told<br />

that this part of the e x pe rim e nt dealt w ith recog nition of stim uli pres ente d<br />

v e ry brie ¯ y . As in the ® rst part of the study , a prim e (happy fac e , ang ry<br />

fac e , or poly g on) w as presente d for 1 0 m se c and w as im m ediate ly m aske d<br />

by a 2 -s ec ond prese ntation of an ide og raph. O ne se c ond afte r the ide og raph<br />

disappe are d, tw o pic tures w e re ¯ ashe d on the sc ree n for 7 s e c onds: An<br />

im ag e of the ac tual prim e on one side of the sc ree n and an alte rnate im ag e<br />

(a dis trac tor) on the othe r s ide of the sc ree n. S ubjec ts indic ate d w hic h of<br />

the tw o im ag e s the y thoug ht w as the prim e by pressing one of the six<br />

buttons on a res pons e box . Three buttons to the le ft indic ate d, w ith de g ree s<br />

of c e rtainty v ary ing from ``sure’ ’ to ``jus t g ue ssing ’ ’ , that a c orrec t pic ture<br />

of the prim e w as on the le ft part of the sc reen. Three buttons to the rig ht<br />

indic ate d, w ith v arious de g ree s of ce rtainty , that the c orrec t prim e w as on<br />

the rig ht. Using this res pons e sc ale , the num be r of response options on the<br />

force d-c hoic e te st w as ide ntic al w ith the num be r of response options in the<br />

liking part. The sc ree n position of the c orrec t im ag e w as random ise d. The<br />

distrac tor for the fac ial prim e w as a fac e w ith the sam e e x pre ssion and the


sam e g ende r, but of a diffe rent pe rson. The dis trac tor for the poly g on w as a<br />

diffe rent poly g on.<br />

This de s ig n of the force d-c hoic e te st, using tw o diffe rent pe ople w ith<br />

the sam e fac ial e x pression, controls for the ability to rec og nise a particular<br />

fac e. How e v e r, it doe s not c ontrol for w he the r subje c ts w ere aw are of the<br />

sm iling v e rsus the ang ry e x press ions us e d as prim es . T hat is , the results of<br />

the force d-c hoic e te st le av e ope n w he the r subjec ts hav e ac c e ss to the v ale nc e<br />

of the prim e . A lthoug h a diffe rent de s ig n c ould he lp answ e r this que stion, the<br />

prim ary purpos e of the force d-c hoic e te st w as to e stablis h w he the r subje cts<br />

c ould pe rce iv e the de sc riptiv e fe atures of the prim e s c onsc ious ly .<br />

Inter v ie w s . Afte r the e x perim ent, subje c ts w e re inte rvie w e d and probe d<br />

for the ir aw are ne ss of the ``othe r pic tures ’ ’ (prim e s), the prese nc e of<br />

affe c tiv e reac tions to prim e s and ide og raphs, and the ir strate g ie s for<br />

m aking the liking judg e m e nt. S ubjec ts w e re also aske d about their aw arene<br />

ss of the prim es during the force d-c hoic e tas k, and the ir strate g ie s for<br />

m aking the force d-c hoic e judg e m e nt. A fte r the interv ie w e ac h subje c t w as<br />

thoroug hly de brie fe d.<br />

Design<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 443<br />

FIG . 2. S e qu enc e of e v e n ts on on e tria l d urin g fo rc e d -c ho ic e tas k.<br />

The judg em e nt part of the study re¯ e c ts a 5 3 4 3 2 m ix e d de sig n. The<br />

Ty pe of Ex pe c tation fac tor w as m anipulate d be tw e e n subje c ts (no e x pe c tations,<br />

nonspe c i® c prim e e x pe c tati on, pos itiv e prim e e x pe ctati on, ne g ative


444 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

prim e e x pe c tati on, positiv e and ne g ative prim e e x pec tation). The T y pe of<br />

Prim e fac tor w as m anipulate d w ithin subjec ts (happy fac e , ang ry fac e ,<br />

poly g on, or nonprim e ). Finally , the prese ntati on order w as repres ente d<br />

by the w ithin-subje c ts fac tor B loc k O rde r (® rst 2 0 trials , last 2 0 trials ).<br />

R esults<br />

The ke y que s tions addresse d by our analy se s are : (1 ) w he the r sublim inal<br />

prim es in¯ ue nc e s ubje cts’ judg e m ents of the ide og raphs; and (2 ) w he the r<br />

this in¯ ue nc e is m odi® e d by subje cts’ e x pe c tati ons about the alle g e d<br />

im pact of sublim inal prim e s. The analy se s g iv e n late r are base d on data<br />

from the ® rst bloc k (2 0 trials). The data from the se c ond bloc k (last 2 0<br />

trials) w e re also analy s e d and w ill be dis c usse d in a late r se c tion. Means<br />

from both bloc ks are prese nted in Table 1 .<br />

Prim in g e ffe c t. A 4 (ty pe of prim e ) 3 5 (ty pe of ex pe c tati on) MA NOV A<br />

w as pe rform e d on the liking judg em e nts. This analy sis reveale d a sig nific<br />

ant m ain e ffe c t for ty pe of prim e [ F (3 ,17 4 ) = 4 .72 , P < .00 5 ] . The uppe r<br />

pane l of Table 1 show s that in eac h c ondition subje c ts rate d the ide og raphs<br />

prec e ded by happy prim e s hig he r than ide og raphs prece de d by ang ry<br />

prim es . More spe c i® c ally , paired t-te sts reveale d that ac ross all c onditions<br />

TA B LE 1<br />

Exp er im ent 1: M ea n Liking of Id eo gra ph s as a Fu nctio n of Prim e, Expectatio n<br />

Prim e<br />

No<br />

Exp e c t.<br />

Co ndition , and T ria l Blo ck<br />

No n sp e c .<br />

Exp e c t.<br />

Exp e rim e n tal C o n d itio n<br />

Po s.<br />

E xp e c t.<br />

Ne g .<br />

Exp e c t.<br />

Po s. & Ne g .<br />

Exp e c t.<br />

F ir st b lo c k<br />

H appy 3 .63 3 .9 8 4 .0 3 3 .5 8 3 .8 5 3 .8 1<br />

A ng ry 3 .47 3 .2 5 3 .4 3 3 .4 3 .6 3 .4 3<br />

Poly g on 3 .46 3 .4 5 3 .7 8 3 .6 3 .6 5 3 .5 8<br />

N o prim e 3 .27 3 .7 2 3 .7 7 3 .8 2 3 .5 7 3 .6 2<br />

Total 3 .46 3 .6 3 .7 5 3 .6 3 .6 7 3 .6 1<br />

Se c o n d b lo c k<br />

H appy 3 .69 3 .3 1 3 .4 3 .6 8 3 .7 2 3 .5 6<br />

A ng ry 3 .74 4 .2 3 .9 3 3 .6 8 3 .9 3 .8 9<br />

Poly g on 3 .46 3 .3 4 3 .7 7 3 .3 5 3 .5 8 3 .5<br />

N o prim e 3 .4 3 .7 1 3 .5 8 3 .2 3 .6 3 3 .5<br />

Total 3 .57 3 .6 4 3 .6 7 3 .4 8 3 .7 1 3 .6 1<br />

n : 1 4 1 3 1 2 1 2 1 2 6 3<br />

No te: S c ale rang e is 1 ± 6 : 1 , don’ t li ke the ide o graph; 6 , li ke the ide o g raph a lo t.<br />

All


SU B LIM IN A L A FFECT IV E P RIM IN G 445<br />

ide og raphs prec e de d by happy prim e s w e re rate d hig he r than ide og raphs<br />

prec e de d by angry prim e s [ t(6 2 ) = 3 .6 2 , P < .0 1 ; poly g ons t(6 2 ) = 2 .6 3 , P <<br />

.02 ; or no prim e s t(6 2 ) = 2 .0 2 , P < .0 5 ] . The se ® nding s represe nt an ov e rall<br />

affe c tiv e prim ing e ffe c t.<br />

Effe c ts o f e xpe c ta tio n s. The MANO V A revealed no s ig ni® c ant prim e 3<br />

ty pe of e x pe c tation inte rac tion, F < 1 , indic ating that the ov erall s tre ng th of<br />

the affec tiv e prim ing e ffe c t did not de pe nd on subje c ts’ e x pe c tations about<br />

the alle g e d im pac t of sublim inal prim e s. More de taile d analy s es c on® rm e d<br />

this c onclus ion. S pe c i® c ally , w e c reate d tw o indic e s of the e ffe ctiv e ne ss of<br />

affe c tiv e prim e s. O ne inde x m e asured the e ffe ctiv e ne ss of positiv e prim e s,<br />

the othe r m easured the e ffe c tiv e ne ss of ne g ativ e prim e s. The s e inde x e s<br />

w e re calc ulate d by taking the differenc e be tw e e n e ach subje c t’ s liking of<br />

ide og raphs prec e de d by affe ctiv e prim e s and this subje c t’ s liking of<br />

ide og raphs prec e de d by nonaffe c tiv e prim e s (poly g ons and blank prim e s).<br />

As show n in Table 2 , the patte rn of m e ans did not c onform to the<br />

predic tions de riv e d from the fe e ling s-as-inform ation ac c ount. The<br />

obtai ne d diffe renc e s in judg em e nts of ide og raphs prec e de d by happy<br />

v e rsus angry prim e s w e re m ost pronounc e d in the nonspec i® c e x pe c tati on<br />

c ondition [ t(1 2 ) = 4 .2 , P < .00 2 ] . Henc e , a m anipulation ex pe c te d to<br />

unde rm ine the e ffe c t of sublim inal prim ing appe ars to hav e stre ng the ne d<br />

it. S im ilarly , the e ffe c t of positiv e prim e s in the positiv e e x pe c tati on<br />

c ondition, w he re the fe e ling -as -inform ation acc ount predicts a disc ounting<br />

of positiv e affe c t, w as nonsig ni® c antly s trong e r than in the ne g ative<br />

Prim e<br />

TA B LE 2<br />

Experim ent 1: Effectiveness of A ffective Prim es as a Fun ction of P rim e,<br />

E xp ectation Con dition, an d Tria l Blo ck<br />

F ir st b loc k<br />

No<br />

Exp e c t.<br />

No n sp e c .<br />

Exp e c t.<br />

Exp e rim e n ta l C o n d itio n<br />

Po s.<br />

Exp e c t.<br />

Ne g .<br />

Exp e c t.<br />

Po s. & Ne g .<br />

Exp e c t.<br />

Happy .2 6 .4 .2 6 2 .1 2 .2 4 .2 1<br />

A ng ry .1 1 2 .3 4 2 .3 4 2 .3 1 2 .0 1 2 .1 7<br />

Se c o n d b lo c k<br />

Happy .2 6 2 .2 2 2 .2 7 .4 1 .1 1 .0 6<br />

A ng ry .3 1 .6 8 .2 6 .4 1 .2 9 .3 9<br />

n : 1 4 1 3 1 2 1 2 1 2 6 3<br />

No te: The e ffe c tiv e ne ss rating of happy and ang ry prim e s re p re se nts the dif fe re nc e<br />

be tw e e n rating s of ide og raphs on tria ls w ith affe c tiv e prim e s and tri als w ith ne utral<br />

prim e s or no prim e s at all Ð e .g . e ff e c tiv e ne ss of happy pri m e = happy 2 (ne utral +<br />

none )/2 .<br />

All


446 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

e x pe ctati on c ondition, w he re an aug m e ntati on e ff ec t w as predic ted [ t(2 2 ) =<br />

1 .4 0 , P < .18 ] . Moreov e r, the e ffe c t of ne g ati v e prim e s in the positiv e<br />

e x pe ctati on c ondition w as not diffe rent from the ir im pac t in the ne g ativ e<br />

e x pe ctati on c ondition [ t(2 2 ) = 0 .1 ] . S im ilarly , the e ffe c ts of both ty pe s of<br />

prim es in the c ondition w he re subjec ts e x pe c te d positiv e as w e ll as<br />

ne g ativ e in¯ ue nc e s did not differ from the no-e x pec tation conditions , in<br />

c ontrast to predic tions from the fe e ling -as -inform ation m ode l. S urpris -<br />

ing ly , in contrast to Murphy and Z ajonc ’ s (1 9 9 3 ) previous ® nding s, in<br />

the no-ex pe c tati ons c ondition, no sig ni® c ant prim ing e ffe c t w as obtai ne d<br />

[ t(1 3 ) = 0 .9 8 ] , althoug h the m eans w e re in the predic te d direc tion.<br />

In com bination, the se ® nding s indic ate that inform ing subje c ts about a<br />

possible in¯ uenc e of sublim inally prese nte d m ate rials did not result in the<br />

predic te d aug m entati on and disc ounting e ffe c ts . If any thing , induc ing thes e<br />

e x pe ctati ons e nhanc e d the im pac t of the prim ing proc edure, in partic ular<br />

w he n the e x pec tations w e re nons pe c i® c . In fac t, affe c tiv e prim ing e ffe c ts<br />

w e re only obtai ne d unde r e x pec tation c onditions and not unde r the noe<br />

x pe ctati on c ondition that replic ate d the proc e dures use d by Murphy and<br />

<strong>Zajonc</strong> (1 9 9 3 ). The latte r obse rvation is inc ons is te nt w ith the predic tion<br />

de riv e d from the affe c tiv e prim ac y ac c ount that the affe ctiv e prim ing is<br />

fully indepe nde nt of hig he r-order inte rv e ntions.<br />

C o rre c tiv e p ro c e s se s. B ase d on the fe e ling s-as-inform ati on approac h,<br />

one m ig ht e x pe c t a ne g ati v e relation be tw e e n the spe e d w ith w hic h subje c ts<br />

form a judg e m e nt and the e ffe c tiv e ne ss of the prim e s. T his w ould re¯ e c t<br />

the possibility that the dis c ounting and aug m enting proc e sse s take tim e , a<br />

possibility that has not be e n addresse d in previous m isattribution studie s.<br />

To te st this c onje c ture, w e c om pute d the c orrelation betw e e n the e ffe c -<br />

tiv e ne ss of prim es (as de® ne d e arlie r) and response tim es (in bloc k 1 ). In<br />

the no-ex pe c tati on c ondition this c orrelation w as [ r(1 4 ) = .43 , n.s.] . The<br />

m e an c orrelation in the four-e x pe c tati on c onditions w as [ r(4 9 ) = .14 , n.s.] .<br />

This ® nding sug g e sts the e ffe c t of the prim es w as indepe nde nt of subje c ts ’<br />

atte m pts to c onside r the inform ati onal v alue of the ir fe e ling s in the c onte x t<br />

of the e x pec tations c onv e y e d to the m .<br />

Affe ctiv e e xp e rie n c es . S ubje c ts w e re aske d if the y e x pe rie nc e d any<br />

affe c tiv e reac tions to the prim e s or the targ e t ide og raphs . The y w e re also<br />

aske d to des c ribe how the y m ade their judg e m e nts of ide og raphs. No<br />

subjec t reporte d e x pe rie nc ing any fe e ling s as a result of the prim ing .<br />

S e v e n subje c ts (1 1 % ) reporte d basing the ir judg e m e nts on the ir ``g utreac<br />

tions’ ’ to the ide og raphs. The s e subje c ts w e re e v e nly distribute d<br />

ac ross e x pe c tati on c onditions and did not sig ni® c antly diffe r from othe r<br />

subjec ts in the ir response to the prim ing m anipulation, F < 1 . O the r<br />

subjec ts responde d that the y base d the ir judg e m e nts on ae sthe tic fe atu res<br />

of the ide og raphs, w hat the y rem inde d them of, or s im ply on how m uc h<br />

the y like d the m . T his ® nding is consiste nt w ith the predic tion of the


SU B LIM IN A L A FFECT IV E P RIM IN G 447<br />

affe c tiv e prim ac y hy pothe sis that affe c tiv e re ac tions e lic ite d by the sublim<br />

inal prim e s are not represe nte d as c ons c ious fe e ling s.<br />

Re su lts o f b lo c k 2 . Follow ing the proc e dures use d by Murphy and<br />

<strong>Zajonc</strong> (1 9 9 3 ), the prese nt e x pe rim ent had tw o bloc ks of trials. In the<br />

se c ond block, the sam e ide og raphs that w e re presente d in the ® rst bloc k<br />

w e re judg e d ag ain by the sam e s ubje cts. How e v e r, the repe ate d ide og raphs<br />

w e re prec e de d by fac ial e x pre ssions of the opposite v ale nc e (i.e. an<br />

ide og raph prim e d w ith a happy fac e in bloc k 1 w as prim ed w ith an ang ry<br />

fac e in block 2 ). In the Murphy and <strong>Zajonc</strong> (1993) e x pe rim e nt, the prim ing<br />

e ffe c t in the se c ond bloc k of trials w as w e ake r c om pare d w ith the e ffe c t in<br />

the ® rst bloc k. This sug g e sts that affe ctiv e prim ing e ffe c ts e s tablishe d<br />

during the ® rst e x posure m ay be quite robust and dif® c ult to oblite rate<br />

by affe c tiv e prim ing during the se c ond e x posure. The refore, data from<br />

bloc k 2 w e re analy se d se parate ly . A 4 (ty pe of prim e ) 3 5 (ty pe of<br />

e x pe c tati on) MA NO V A w as pe rform e d on the liking judg e m e nts from<br />

the se c ond bloc k. This analy sis reveale d only a sig ni® c ant m ain e ffe c t<br />

for ty pe of prim e [ F (3 ,1 7 4 ) = 6 .6 7 , P < .0 0 1 ] , and no inte rac tions . The<br />

patte rn re¯ e c te d in this m ain e ffe c t, how e v er, w as o p p o s ite to the patte rn<br />

obse rv e d for the ® rst bloc k. S pe c i® c ally , paired t-te sts rev eale d that ac ross<br />

all c onditions ide og raphs assoc iate d w ith happy prim e s in bloc k 2 w e re<br />

like d le ss than ide og raphs ass oc iate d w ith ang ry prim e s in bloc k 2 [ t(6 2 ) =<br />

2 .8 7 , P < .01 ] . A ll diffe renc e s be tw e e n ty pe s of prim es m irrored the e ffe cts<br />

obtai ne d in bloc k 1 . Ideog raphs prec e de d by happy prim e s w e re not like d<br />

les s than ide og raphs prece ded e ithe r by poly g ons or no-prim e s (t < 1 ),<br />

w he reas ide og raphs prec e de d by ang ry prim e s w e re like d m o re than<br />

ide og raphs prec e de d by poly g ons, or no-prim e s [ t(6 2 ) = 4 .8 6 , P < .00 1 ] .<br />

This patte rn presum ably re¯ e c ts that the ide og raphs be c am e ass oc iate d<br />

w ith an affe c tiv e e v aluation during the ® rst e x posure. In othe r w ords, in<br />

bloc k 1 s ubje cts m ay hav e form e d and retai ne d an atti tude tow ard the<br />

ide og raph. T his e v aluation m ay hav e de te rm ine d judg e m e nts during the<br />

se c ond e nc ounte r w ith the ide og raph, w hate v e r the prim e that prec ede d the<br />

e x pos ure in the se c ond bloc k. If this e x planation is c orrec t, one should<br />

e x pe c t a positiv e c orrelati on be tw e en the ® rst judg e m e nt and se c ond<br />

judg e m ent of the sam e ide og raph, de spite pairing w ith opposite prim e s.<br />

In fac t, judg em e nts of repeate d ide og raphs w e re positiv e ly and sig nific<br />

antly c orrelate d [ r(6 3 ) = .44 , P < .0 0 1 ] . A s a c ontrol w e also c om pute d the<br />

c orrelati on be tw e e n c om parable judg e m e nts of nonrepe ate d ide og raphs.<br />

The se judg e m e nts w e re unc orrelate d [ r(6 3 ) = .0 0 4 , n.s.] . The diffe renc e<br />

be tw e e n the aforem e ntione d c orrelation c oe f® c ients w as sig ni® c ant, z =<br />

2 .5 6 , P < .05 .<br />

Treating bloc k as a w ithin-subje c ts fac tor, w e obtai n a tw o-w ay inte rac<br />

tion of ty pe of prim e and bloc k orde r [ F (3 ,17 4 ) = 8 .3 7 , P < .0 0 1 ] . This<br />

re¯ e c ts that the e v aluations are c onsiste nt w ith the v ale nc e of the prim e in


448 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

bloc k 1 [ F (3 ,1 7 4 ) = 4 .7 2 , P < .0 0 5 ] for the s im ple m ain e ffe c t, but<br />

inc ons is te nt w ith the v ale nc e of the prim e in bloc k 2 [ F (3 ,1 7 4 ) = 6 .67 ,<br />

P < .0 0 1 ] for the sim ple m ain e ffe c t.<br />

F o r c e d -c h o ic e te s t. Data from the force d-c hoic e te st of prim e rec og nition<br />

indic ate that the sublim inal pre se ntati on proc e dures w e re suc c e ss ful.<br />

The rec og nition ratio (i.e . the num be r of c orrec t rec og nitions of the prim e<br />

div ide d by the total num ber of rec og nition judg e m ents m ade) w as .48 ,<br />

S D = .1 0 , w hic h is not sig ni® cantly diffe rent from the c hanc e e x pe c tation<br />

of .5 0 . T he rec og nition ratio did not diffe r as a func tion of ty pe of prim e<br />

(happy fac e , ang ry face , poly g on), e x pe rim e ntal c ondition, and prac tic e<br />

(® rst 1 5 v s. last 1 5 judg em e nts). There w as no relationship be tw ee n<br />

subjec ts’ pe rform anc e on the force d-c hoic e tas k and the e ffe c tiv e ne ss of<br />

prim es in the judg e m e nt tas k.<br />

Discussion<br />

In sum m ary , an affe c tiv e prim ing e ff ec t ac ross v arious instruc tional c onditions<br />

w as obtai ne d in the ® rst bloc k of trials . This ® nding prov ide s a<br />

g e ne ral replic ation of the phe nom e non found by Murphy and Z ajonc<br />

(1 9 9 3 ) in study 1 , althoug h the results obtained in the no-e x pe c tati ons<br />

c ondition, w hic h prov ide d a direc t replic ation of Murphy and Z ajonc’ s<br />

proc edures, w e re not s ig ni® c ant.<br />

In addition, the results of the se c ond bloc k of trials sug g e st that affe c tiv e<br />

prim ing has a las ting e ffe c t on prefe renc e s. Judg em e nts of the ide og raphs<br />

in the se c ond bloc k w e re positiv ely correlate d w ith judg em e nts of the sam e<br />

ide og raphs in the ® rst bloc k (de spite be ing paired w ith oppos ite prim e s.)<br />

W e sug g e st that this patte rn represe nts an e ffe c t obse rve d in studie s that<br />

found that a sing le pairing of an affe c tiv e prim e w ith a neutral targ e t w as<br />

suf® c ie nt to bias the subse que nt im pression form e d of that targ e t (Nie -<br />

de nthal, 1 9 9 0 ; Kros nic k, B e tz , Jussim , & Ly nn, 1 9 9 2 ). This ® nding also<br />

sug g e sts that af fe c tiv e prim ing m ay be e spe c ially e ffe c tiv e at the initial<br />

attitude -form ation stag e and m ay be le ss e ff ec tiv e at the atti tude -c hang e<br />

stag e . O n a the oretic al le v e l, lite rature on autom atic e v aluation propose s<br />

that a ne utral obje c t c an acquire an e v aluati on as a result of pairing w ith an<br />

e v aluativ e obje c t (B arg h e t al., 1 9 9 2 ; Faz io, S anbonm ats u, Pow e ll, &<br />

Karde s, 1 9 8 6 ). O nce the obje c t has its ow n e v aluation assoc iate d w ith it,<br />

it w ill be les s susc e ptible to s ubse que nt prim ing . This inte rpretation<br />

ass um e s, of c ourse , that the prim ing m anipulation c ontinue d to be as<br />

e ffe c tiv e in the se c ond bloc k. 1<br />

1 W e thank an anony m ous re v ie w e r for sug g e sting this interpre tation.


SU B LIM IN A L A FFECT IV E P RIM IN G 449<br />

Most im portant, how ev e r, the abs enc e of attribution e ffe c ts sug g e sts that<br />

the prim ing effe c t w as m e diate d by a proc e ss othe r than that sug g es te d by<br />

the fe e ling s-as-inform ation approach. S ubje c ts w ho w e re told that the<br />

unc onsc ious prim e s m ig ht affe c t the ir judg e m e nts w e re as susc e ptible to<br />

the ir in¯ ue nc e as uninform ed s ubje c ts and ne ithe r a disc ounting e ffe c t nor<br />

an aug m entati on e ffe c t w as obse rve d. On the c ontrary , the judg e m e nts<br />

prov ide d by subje c ts w ho w e re inform e d that the y w ould be e x pos ed to<br />

sublim inal prim e s, but w e re not inform e d about the spe c i® c nature of the ir<br />

im pac t, show e d the strong e st prim ing e ffe c t. Note , how e v er, that this latte r<br />

® nding is also inc onsiste nt w ith the null e ffe c t predic tions g e nerate d by the<br />

affe c tiv e prim ac y hy pothe sis. As a p o st-h o c e x planati on, w e sug g e st that<br />

the nonspe c i® c ex pe c tati on m anipulations inc rease d s ubje cts’ atte ntion to<br />

the dis play , thus inc reasing the im pac t of sublim inal prim e s. T he fac t that<br />

the spe e d w ith w hic h subje c ts prov ide d a judg e m e nt w as unrelate d to the<br />

e ffe c tiv e ne ss of the prim e s furthe r sug g e sts that the im pac t of the prim e s<br />

w as inde pe nde nt of subje c ts ’ atte m pts to relate the ir affe c tiv e response to<br />

the situational in¯ ue nc e s de sc ribe d to the m . Finally , subje c ts’ s e lf-reports<br />

indic ate d that the y w e re not c ons ciously aw are of any fe e ling reac tions to<br />

the prim es or the targ e ts.<br />

In g e ne ral, tw o e x planati ons m ay be offe red for the ® nding s of Ex pe rim<br />

e nt 1 . On one hand, the data are c ons iste nt w ith the predic tion of the<br />

affe c tiv e prim ac y ac c ount, w hic h holds that the affe c tiv e sy s te m is im pene -<br />

trable to m anipulations of infe renc e and that it inv olv e s autom ati c proc<br />

e sse s for w hic h subje c ts are not able to c orrec t. Moreov e r, as predic te d,<br />

sublim inal prim ing produc e d affe c t that w as not represe nte d as a fe e ling .<br />

From this pe rspe c tiv e , the m isattribution m anipulati on w as unsuc c e ssful<br />

be c ause subje c ts c ould not c orrec t the ir judg em e nts for an in¯ ue nc e the y<br />

did not pe rce iv e . O n the othe r hand, it is c onc e iv able that our m isattribution<br />

m anipulati on w as not suc c e ss ful for m e thodolog ic al reas ons. R oss and<br />

Olson (1 9 8 1 ) disc usse d v arious reasons w hy m isattribution m anipulati ons<br />

m ay fail. T w o reasons that apply to our paradig m are : (1 ) the plausibility of<br />

the m isattributional source as a c aus e of the e x pe rie nc e ; and (2 ) the<br />

salie nc e of the m isattribution source at the tim e of the e x pe rie nc e . Re g arding<br />

plaus ibility , our subje c ts w e re told that affe c tiv e c hang e s w ould be<br />

induc e d by sublim inal prim e s. A lthoug h the c ause of affe c t g iv e n to our<br />

subje c ts w as c le arly the true c ause , subje c ts m ay not hav e belie v e d that<br />

sublim inal prim es w e re prese nte d, or w e re e ffe c tiv e, and m ay sim ply hav e<br />

disc arde d this inform ation as bog us . Alte rnatively , our subje c ts m ay hav e<br />

be lie v e d w hat the y w e re told, but bec aus e the prim es w e re inv isible , the y<br />

m ay hav e quic kly forgotte n about the ir ex is te nc e , re¯ e cting a low salie nc e<br />

of the m isattribution m anipulation. Moreov e r, subje c ts m ay hav e disreg<br />

arde d the attributional inform ation as irrele v ant be c ause the y w e re not<br />

c onsc ious of e x pe rie nc ing affe c tiv e reac tions. Finally , e v e n if s ubje cts


450 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

discerned their a ffective reactions, they m ay h ave felt con fused o r overwhelmed<br />

by the a ttributio nal arithme tic required fro m them . F or ex am ple, in the positive<br />

e x pe ctati ons c ondition, the tas k required subje c ts to trust ne g ativ e reac tions,<br />

but to distrust positiv e reac tions. Giv e n that the v ale nc e of the affe c tiv e<br />

prim e c hanged from trial to trial, as the orde r of happy and ang ry prim e s w as<br />

random , ke e ping track of all of the reac tions m ay hav e be en too dif® c ult.<br />

EXPER IM ENT 2<br />

Giv e n the se c onc e rns, w e c onduc ted a s e c ond study , base d on a m odi® e d<br />

prim ing m anipulation and a diffe rent, m ore s alie nt, and m ore plausible<br />

m isattribution m anipulation. In redes ig ning both the prim ing m anipulation<br />

and the attribution m anipulati on w e trie d to ac hie v e m ax im um sim plic ity<br />

and phe nom enal c larity for s ubje c ts . First, if a dis ting uishable fee ling is<br />

produc e d by s ublim inal m anipulations, in c ontrast to w hat the affe c tiv e<br />

prim ac y hy pothe sis w ould sug g e st, w e w ante d to m ake it v e ry e as y for<br />

subjec ts to recog nise it. A c c ording ly , e ac h se t of trials w ith affe c tiv e prim e s<br />

w as prec e de d by a se t w ith ne utral prim e s to fac ilitate rec og nition of any<br />

affe c tiv e re ac tions. Moreov e r, se que nc e s of ® v e affe ctiv e prim es of the sam e<br />

v ale nc e w e re presente d to prolong and inte nsify affe c tiv e reac tions . S e c ond,<br />

w e use d a m isattribution m anipulati on that has be e n s uc c e ssfully e m ploy e d<br />

in pre v ious studie s (S c hw arz e t al., 1 9 9 1 ). S pe c i® c ally , w e play e d proc e sse d<br />

Ne w A g e m us ic throug hout the e x pe rim e nt and inform ed subje c ts that this<br />

m usic w ould e lic it positiv e (or ne g ative) fe e ling s. T o ac c om m odate the<br />

initial ne utral trials, subje c ts w ere furthe r inform ed that the se fe e ling s m ay<br />

only se t in afte r som e tim e, thus allow ing the m to attribute any affe c tiv e<br />

e x pe rie nc e s e lic ite d by the s ubse que nt affe c t trials to the im pac t of the m usic .<br />

In c om binati on, the se c hang e s s hould fac ilitate the de te c tion of affe c tiv e<br />

c hanges, inc reas e the salie nc e of the ir alle g e d alte rnati v e source , and m otiv<br />

ate subje c ts to e ng ag e in appropriate attributional proc e sse s. A s in Ex pe rim<br />

e nt 1 , the fee ling s-as -inform ation m ode l predic ts dis c ounting e ffe c ts on<br />

trials w he re the aff ec tiv e e x pe rie nc e m atc he s the alle g e d im pac t of the<br />

m usic , and aug m entati on e ffe c ts on trials w here the affec tiv e e x pe rie nc e<br />

c ontradic ts the alle g e d im pac t of the m usic. In c ontrast, the affe c tiv e prim ac y<br />

m ode l predic ts no im pact of the attributional m anipulati ons.<br />

M ethod<br />

Subjects<br />

Eig hty -se v en m ale and fe m ale unde rgraduate subje c ts at the Univ e rsity of<br />

Mic hig an partic ipate d in partial ful® lm e nt of a c ourse requirem e nt. S ubje c ts<br />

w ho kne w C hine se , Korean, or Japane se w e re e x cluded from partic ipation.


M aterials, Apparatus, and Procedure<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 451<br />

Prim in g m a n ip u la tio n . S ubje c ts w e re ag ain aske d to g iv e liking judg e -<br />

m e nts of ide og raphs , us ing the sam e m e thod as in Ex pe rim ent 1 . Half of the<br />

ide og raphs w e re prec e de d by fac ial affe c t prim e s (happy and ang ry), and<br />

half w e re prec e de d by ne utral prim e s. In contras t to Ex pe rim e nt 1 , the<br />

ne utral prim e s w e re not poly g ons , but fac e s of indiv iduals w ho w e re aske d<br />

to pose not show ing any spe ci® c e m otion. This c hang e e lim inate s the<br />

c om parability proble m be tw e e n affe c tiv e ly c harg e d prim es and ne utral<br />

prim e s. In Ex pe rim e nt 2 , all prim es w e re fac ial photog raphs that diffe red<br />

only on the ir aff ec tiv e dim e nsion. The e x posure, lig hting , distanc e , and<br />

response c onditions w e re ide ntical to Ex perim e nt 1 .<br />

Or d e r o f stimu li. The proc e dure starte d w ith ® v e w arm -up ide og raphs<br />

prec e de d by ne utral prim e s. The te st ideog raphs w e re prec ede d by four<br />

g roups of ® v e prim e s that alte rnate d be tw e e n ne utral and affe c tiv e fac e s.<br />

The ® rst g roup of prim e s c onsiste d of ne utral fac e s, the se c ond of affe c tiv e<br />

fac es , the third of ne utral fac e s, and the las t of affe c tiv e fac e s ag ain. The<br />

ty pe of aff ec tiv e prim e (happy or ang ry) follow ing the ® rst g roup of ne utral<br />

prim e s w as c ounterbalanc e d, c reating tw o orde rs of stim uli. This c ounterbalanc<br />

ing allow s the e x am ination of the im pac t of affe c tiv e response s that<br />

are consiste nt or inc ons is te nt w ith e x pe c tati ons induc e d by the attributional<br />

m anipulations. A fte r the bloc k of 2 0 trials w as c om ple te d, it w as repeate d<br />

w ith a ne w se t of s tim uli to e x plore c hang es in affe c tiv e prim ing m anipulation<br />

ov e r tim e . Unlike E x pe rim ent 1 , how ev e r, stim uli use d in bloc k 2<br />

w e re not repe titions of stim uli us e d in bloc k 1 .<br />

Ma n ip u la tio n o f in fer e n c e s. W hile subje c ts rate d the ide og raphs , e m otionally<br />

am big uous Ne w A g e m usic recorde d at half-s pe e d w as play e d in<br />

the bac kg round. This m usic w as c le arly audible from the be g inning of the<br />

rating task, inc reas e d in loudne ss for the ® rst 3 0 se c onds, and c ontinue d for<br />

W arm -up trials Test Trials<br />

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -<br />

- - NN NN N - - - - NN NN N - - H HHH H - - NN NN N - - AAAA A - -<br />

- - NN NN N - - - - NN NN N - - A AAA A - - NN NN N - - HHHH H - -<br />

Note:<br />

N ± neutral facial expression prim e<br />

H ± happy facial expression prim e<br />

A ± angry facial expression prim e<br />

FIG . 3. E xp e rim e n t 2: A rrang e m e n t o f sub lim in al fac ial prim es p re se nted b efo re th e ide o graph s.


452 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

the re st of the e x pe rim ent. S ubje c ts w e re told that althoug h this m us ic w as<br />

not rele v ant for the ide og raph-rating tas k, it w as play ed be c ause w e w e re<br />

inte rested in its de lay e d e ffe c ts on som e tas ks follow ing the ide og raph<br />

rating s. All s ubje cts w e re told that the m usic had be e n show n to hav e an<br />

e m otional im pac t on m ost pe ople , but that it w as unknow n how long it<br />

w ould tak e for the m usic to hav e an im pac t, only that this im pac t w ould not<br />

be im m ediate . To ke e p the subje c ts v ig ilant about the m us ic -relate d<br />

c hanges in the ir affe c tiv e s tate s, all subjec ts w e re as ke d four tim e s<br />

throug hout the e x pe rim e nt: ``Do y ou fe e l af fe c te d by the m usic? ’ ’ The<br />

onse t of this que stion w as tim ed to prec e de shortly be fore the onse t of<br />

affe c tiv e prim e s, thus rende ring the m usic hig hly acc es sible as a possible<br />

source for any reac tions that m ay be e licite d by the prim e s. A s a m anipulation<br />

of ex pe c tati ons, half of the subje c ts w e re inform e d that afte r a w hile<br />

this m usic w ould m ake the m fe e l ple asant, w he reas the other half w e re<br />

inform e d that the m us ic w ould m ake the m fe e l unple as ant.<br />

F o r c e d c h o ic e . A fte r the judg e m e nt tas k, w e te ste d w he the r the prim e s<br />

w e re prese nte d at a suboptim al le v e l. The proc e dure w as a sim pli® e d<br />

v e rsion of the force d-c hoice tas k from Ex pe rim e nt 1 . S ubje c ts w e re prese<br />

nte d w ith 2 0 trials on w hic h ide og raphs w e re prec e de d by affe c tiv e<br />

prim es . S tim uli and e x posure c onditions w ere ide ntical to the ® rst part of<br />

the study . Te n ide og raphs w ere prec e de d, in random orde r, by happy face s<br />

and te n w e re prec e de d by angry face s. Im m e diate ly after the prim e d<br />

ide og raph w as show n, tw o fac e s appeared on the rig ht and le ft side s of<br />

the sc ree n. O ne face w as an e x ac t c opy of the prim e , the othe r w as a<br />

photog raph of a diffe rent indiv idual of the sam e g e nde r show ing the sam e<br />

fac ial e x pression as a prim e . S ubjec ts w e re aske d to indic ate (by pressing a<br />

button) w hic h part of the sc ree n s how e d the fac e ¯ ashe d just before the<br />

ide og raph.<br />

Inte rv ie w s. A fte r the m ain part of the e x pe rim e nt, w e aske d s ubje c ts<br />

se v e ral que stions about the nature of the ir af fe c tiv e and nonaffe c tiv e<br />

reac tions to the m usic , ide og raphs, and prim e s as w e ll as to the tim ing of<br />

the se reac tions. S ubje c ts w e re also aske d w he the r the y atte m pte d to c orrec t<br />

the ir judg e m ents or w e re aw are of the prim e s, and w e re probe d for suspic<br />

iousne ss. Finally , all s ubje c ts w e re thoroug hly de brie fe d and thanke d.<br />

Design<br />

In c om bination, the se m anipulati ons resulte d in a 3 (happy v s . ne utral<br />

v s . ang ry prim e s) 3 2 (positiv e v s. ne g ative e x pe c tati on about the im pac t<br />

of m usic ) fac torial de sig n, w ith the ® rst fac tor m anipulate d w ithin and the<br />

se c ond fac tor m anipulate d be tw e e n s ubje c ts .


Results<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 453<br />

The ke y que stions addresse d by our analy se s are : (1) w hethe r sublim inal<br />

prim e s in¯ ue nc e subje c ts’ judg e m ents of the ideog raphs; and (2) w hethe r<br />

this in¯ ue nc e is m odi® e d by the attributional m anipulati ons. 2<br />

Prim in g effe c t. A 3 (ty pe of prim e ) 3 2 (ty pe of e x pe c tation) MANO V A<br />

w as pe rform e d on the liking judg e m ents. This analy sis rev e ale d a sig ni® c ant<br />

m ain e ffe c t for ty pe of prim e [ F (2 ,1 7 0 ) = 3 .8 8 , P < .03 ] . A s in Ex pe rim e nt 1 ,<br />

ide og raphs prec e de d by happy prim e s w e re like d sig ni® c antly m ore than<br />

ide og raphs prec e ded by ang ry prim e s, ac ros s both e x pe c tati on c onditions<br />

(se e Table 3 ).<br />

Effe c ts o f e xpe c ta tio n s . T he MA NOV A also revealed a nonsig ni® c ant<br />

m ain e ffe c t of e x pe c tations on judg e m e nts [ F (1 ,8 5 ) = 2 .9 7 , P < .0 9 ] .<br />

W he n subje c ts e x pe c te d the m usic to m ake the m fe e l g ood, the y like d<br />

the ide og raphs m ore than w he n the y e x pe c te d the m usic to m ake the m<br />

fe e l bad. The se e x pec tanc y -c ong rue nt e v aluations rese m ble a plac e bo<br />

e ffe c t (se e Ross & O ls on, 1 9 8 1 ) and are inc om pati ble w ith the infe re<br />

ntial predic tions de riv e d from the fe eling s -as -inform ation m ode l, as<br />

w e ll as w ith the null e ffe c t prediction de riv e d from the affe c tiv e<br />

prim ac y hy pothe sis.<br />

This m ain e ff ec t w as quali® e d by a nonsig ni® c ant prim e 3 ty pe of<br />

e x pe c tati on inte rac tion [ F (2 ,1 7 0 ) = 2 .4 6 , P < .0 9 ] . This inte raction<br />

sug g e sts that the stre ng th of the prim ing e ffe c t de pe nde d on subjec ts’<br />

e x pe c tati ons about the alle g e d im pac t of the m us ic . More de taile d analy<br />

se s of this inte raction reveale d that the patte rn is inc ons iste nt w ith<br />

predic tions of the fe e ling -as-inform ati on m ode l. As in Ex perim e nt 1 , w e<br />

c reate d se parate indic e s of the e ffe c tiv e ne s s of positiv e and ne g ative<br />

prim e s. The se indic e s represe nt the diffe renc e be tw e e n the liking of<br />

ide og raphs prec e de d by the respe c tiv e affe c tiv e prim es and ide og raphs<br />

prec e de d by ne utral prim e s.<br />

As show n in Table 4 , the e ffe c t of the pos itiv e prim e s w as som e w hat<br />

strong e r w he n subje c ts e x pe c te d a positiv e in¯ ue nc e of the m us ic than<br />

w he n the y e x pe c ted a ne g ative in¯ uenc e (t < 1 ). Moreov e r, the e ffe c t of the<br />

2 All data in the re su lts se c tio n of Ex pe rim e nt 2 c om e from the ® rst bloc k of trials . The<br />

se c ond bloc k did not sh ow any m ain e f fe c ts or interac tion e f fe c ts of prim ing and e x pe c tation<br />

on judg e m e nts of ide og raphs. Unli ke Ex pe rim e nt 1 , no stim uli w e re re pe ated in the se c ond<br />

bloc k of Ex pe rim e nt 2 . The se c ond bloc k re pre se nted a sim ple c ontin uation of e v e nts from<br />

the ® rst bloc k w ith ne w pri m e s and targ e ts. The abse nc e of the prim ing e ffe c t p re sum ably<br />

re ¯ e c ts hab ituatio n to the affe c tiv e prim ing proc e dure .<br />

B e c ause there w as no e ffe c t of orde r (i.e . w he ther the tone of the ® rst g roup of affe c tiv e<br />

prim e s w as c onsis tent or inc onsis tent w ith the affe c tiv e tone of the e x pe c tations), data ac ross<br />

both orde r c onditio ns w e re c ollap se d.


454 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

TA BLE 3<br />

Experim ent 2: M ea n Lik in g o f<br />

Id eo gra ph s as a Fun ction o f Prim e<br />

and E xp ectation Co ndition<br />

Prim e<br />

Exp e rim e n ta l C o n d itio n<br />

Po s.<br />

Expe c t.<br />

Ne g .<br />

Exp e c t.<br />

Bo th<br />

Happy 3 .8 2 3 .7 0 3 .7 6<br />

Ne utral 3 .6 9 3 .6 1 3 .6 5<br />

A ng ry 3 .7 5 3 .3 6 3 .5 5<br />

Total 3 .7 5 3 .5 5 3 .6 5<br />

n : 4 4 4 3 8 7<br />

No te: S c ale ra ng e is 1 ± 6: 1 , d on’ t<br />

like the ide og raph; 6 , lik e the ide og<br />

ra p h a lo t.<br />

TA B LE 4<br />

E xp erim en t 2: Effectiven ess o f<br />

A ffective P rim es as a Fun ction o f<br />

P rim e and E xp ectatio n Co ndition<br />

Prime<br />

Exp e rim e n ta l C o n d itio n<br />

ne g ativ e prim es w as sig ni® c antly w e a ke r w he n subje c ts e x pec te d a positiv e<br />

in¯ ue nc e of the m usic than w hen the y e x pe c te d a ne g ative in¯ ue nc e [ t(8 5 )<br />

= 2 .13 , P < .05 ] . The se patte rns are opposite to the disc ounting and<br />

aug m e ntati on e ffe c ts predic te d unde r the s e c onditions. He nc e , for both<br />

se ts of prim e s, m anipulations de sig ne d to elic it a dis counting of any<br />

affe c tiv e reac tion to the prim e am pli® e d the prim es ’ im pac t, w hereas<br />

m anipulations de sig ne d to aug m e nt any affe c tiv e reac tion to the prim e s<br />

w e ake ne d the prim e s’ im pac t, in c ontras t to predic tions of the fe eling s -as -<br />

inform ation m ode l.<br />

A dditional analy s es te ste d w he the r the e ffe c ts of prim ing de pe nde d on<br />

subjec ts’ ac tual e m otional response s to the m usic , as rev eale d by the ir se lfreports<br />

at the e nd of the e x perim e nt. The m ean liking judg e m e nts broke n<br />

dow n by the ty pe of response are show n in Table 5 . A 3 3 3 m ix e d<br />

MA NOV A w ith prim e (happy , neutral, ang ry ) and se lf-reporte d respons e<br />

to the m us ic (pos itiv e , none , ne g ativ e) faile d to reveal a sig ni® c ant prim e<br />

by m usic respons e inte raction [ F (1 5 8 ,4 ) < 1 .1] . How e v e r, furthe r analy se s<br />

reveale d that a sig ni® c ant affe c tiv e prim ing e ff ec t w as obtaine d only w he n<br />

subjec ts reporte d a neg ative e m otional im pac t of the m usic [ t(4 9 ) = 3 .07 ,<br />

P < .01 ] . The fe e ling s-as-inform ati on m ode l cannot ac count for the ov e rall<br />

patte rn of ® nding sÐ the abs enc e of prim ing in the positiv e and no-respons e<br />

to m usic g roups, g iv e n a strong prim ing e ffe c t in the ne g ativ e respons e<br />

g roup. This ® nding w as also not predic te d by the affe c tiv e prim ac y<br />

ac c ount. O ne interpretati on of this result sug g e sts that it ag ain m ay refle c t<br />

Po s.<br />

E xp e c t.<br />

Ne g .<br />

Exp e c t.<br />

Bo th<br />

Happy .1 2 .0 9 .1 1<br />

A ng ry .0 5 2 .2 5 2 .0 9<br />

n : 4 4 4 3 8 7<br />

No te : The e ff e c tiv e ne ss rating o f<br />

h app y an d an g ry pri m e s re p re s e n ts<br />

th e d i f f e re n c e b e tw e e n ra ti n g s o f<br />

ide o g raph s o n trials w ith af f e c tiv e<br />

pri m e s and tri als w ith ne utral pri m e s<br />

(e .g . e ffe c tiv e ne ss of happy prim e =<br />

hap py 2 ne utral) .


SU B LIM IN A L A FFECT IV E P RIM IN G 455<br />

TA B LE 5<br />

Experim en t 2: M ean Likin g of Ideograp hs as a<br />

Function o f Prim e, an d R ep orted Em otio nal<br />

R ea ctio n to T he M usic<br />

Re po rted Em o tio n a l R ea c tio n<br />

Prim e Po s. Mixe d No n e Ne g . All<br />

Happy 3 .6 7 3 .5 6 3 .85 3 .7 8 3 .7 6<br />

Ne utral 3 .7 1 3 .2 4 3 .81 3 .6 2 3 .6 5<br />

A ng ry 3 .6 5 3 .4 4 3 .82 3 .4 4 3 .5 5<br />

Total 3 .6 8 3 .4 1 3 .83 3 .6 1 3 .6 5<br />

n : 1 5 5 1 7 5 0 8 7<br />

No te : S c ale rang e i s 1 ± 6 : 1 , do n ’ t l ike the<br />

ide ograph; 6 , like the ide og raph a lo t.<br />

an atte ntional phe nom e non. A s obse rve d in other dom ains (se e S c hw arz ,<br />

1 9 9 0 ; S c hw arz & Clore, 1 9 9 6 for revie w s), neg ative affe ct m ay sig nal a<br />

proble m atic s ituati on, resulting in inc rease d atte ntion and m otiv ation. If so,<br />

inc reas e d atte ntion m ay ag ain hav e inc rease d the e ffe c tiv e ne s s of sublim -<br />

inal prim e s, as sug g e ste d by the ® nding s of Ex pe rim e nt 1 .<br />

Co rr e c tiv e p ro c e sse s . The fe eling s -as -inform ation m ode l sug g e s ts a<br />

ne g ativ e relati on be tw ee n the res pons e tim es and the e ffe c tiv e ne s s of prim -<br />

ing , be c aus e disc ounting and aug m e nting proc e s se s tak e tim e . A naly se s of<br />

c orrelati ons rev eale d that no suc h relation w as prese nt in the data [ r(8 7 ) =<br />

2 .0 4 , n.s.] . This result s ug g e sts that the e ffe c t of the prim e s w as inde pe nde nt<br />

of w he ther subjec ts e ng ag ed in attributional infe renc e proc e s se s.<br />

In addition, s ubje cts w e re as ke d the follow ing que stion: ``Did y ou try<br />

to c orrec t for the pos sibility that y our reac tions to the ide og raphs m ig ht<br />

be in¯ ue nc e d by y our reac tion produce d by the m us ic ? ’ ’ Forty -one (4 7 % )<br />

of the subje c ts answ e red ``y e s’ ’ . A MANO V A inc luding all rele v ant<br />

e x pe rim e ntal fac tors and the c orrec t fac tors (y e s/no) rev e ale d ne ithe r a<br />

m ain e ffe c t nor any inte ractions, all F s < 1 , inv olv ing the c orrec tion fac tor.<br />

Affe c tive e xpe rie n c e s. In post-e x pe rim e ntal inte rvie w s, 7 0 (8 0 % ) subjec<br />

ts reporte d hav ing e x pe rie nc e d an e m otional reac tion to the m usic<br />

during the judg em e nt task. No s ubje ct m entione d any e m otional e x pe rie<br />

nc e s that c ould be relate d to s ublim inal affe c tiv e prim e s.<br />

F o rc e d -c h o ic e te s t. To e stablis h w he the r subje c ts c ould disting uish<br />

fac ial prim e s presente d sublim inally from distrac tors, w e c alc ulate d the<br />

ratio of c orrect rec og nitions of the fac ial prim e to the total num be r of<br />

rec og nition judg e m e nts . The res ulting ratio of .5 0 (S D = .12 ) w as rig ht at


456 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

c hance e x pe c tati on. S ubje c ts ’ pe rform anc e did not im prov e w ith prac tic e ,<br />

did not de pe nd on w hat the ty pe of prim e w as be ing rec og nise d, and did not<br />

predic t e ffe c tiv e ne ss of prim e s in the judg e m e nt tas k.<br />

Discussion<br />

The results of Ex pe rim ent 2 replic ate d the ke y ® nding s of Ex perim ent 1 .<br />

First, the y replic ate d a g e ne ral sublim inal affe c tiv e prim ing e ffe ct, us ing a<br />

diffe rent arrang em e nt of stim uli. S e c ond, m anipulations of the pe rceiv e d<br />

diag nostic ity of one’ s affe c tiv e reac tion to the prim e s did not result in the<br />

disc ounting and augm e ntation e ffe c ts predic te d by the fe e ling s-as-inform ation<br />

m ode l. Inste ad, the obtaine d patte rns re¯ e c te d ex pe c tati on-c ong rue nt<br />

e v aluations. T he re sis tanc e of sublim inal prim ing to the attributional<br />

m anipulations use d in Ex pe rim e nt 2 is e spe cially m eaning ful bec aus e the<br />

disc ounting m anipulation w as ac c e ssible to subje c ts’ aw are ne ss (i.e. the<br />

m usic w as audible at all tim e s), and w as regarde d by m os t subje c ts as<br />

hav ing an e m otional e ffe c t. Third, the re w as no e v idence for a m e diating<br />

role of attributional proc e sse s w ith regard to the stre ng th of the obtai ne d<br />

prim ing e ffe c ts. Finally , subje c ts did not report c onsc ious ly e x pe rie nc ing<br />

affe c tiv e reac tions relate d to the prim e s, but they did notic e affe c tiv e<br />

c hanges in respons e to the m usic .<br />

GENERAL DISC USSIO N<br />

O v e rall, the results of both e x pe rim e nts show a form of affe c tiv e in¯ ue nc e<br />

diffe rent from that e x plaine d by the fe e ling s -as -inform ation m ode l. S ubje<br />

c ts ’ judg em e nts did not show the e x pe c te d patte rn of disc ounting and<br />

aug m e ntati on, and the prim ing m anipulation did not e lic it c onsc ious fe e ling<br />

s that subje c ts c ould report. B e fore w e c onside r the im plic ation of thes e<br />

results, how e v er, w e ne e d to address som e m e thodolog ic al issue s and<br />

c av e ats .<br />

Pow er<br />

The abse nc e of the attributional e ffe c ts predic te d by the fe e ling s -as-inform<br />

ation m ode l is a null ® nding . A c c e pting a null ® nding requires that the<br />

studie s g av e the e ffe c t a reasonable chanc e to m anife st itse lf. C onc e rns<br />

about pow e r are furthe r am pli® e d by the fact that the (m is )attribution<br />

e ffe c ts w e re tes te d be tw e e n subje c ts, w he reas the prim ing e ffe c ts w e re<br />

te ste d w ithin subje c ts. A lthoug h the se de s ig n fe atu res prov ide d m ore<br />

pow e r for de te c ting affec tiv e prim ing e ffe c ts than for de te c ting attribu-


SU B LIM IN A L A FFECT IV E P RIM IN G 457<br />

tional e ffe c ts, se v e ral reasons m ake us doubt that the c ruc ial null ® nding is<br />

due to a lac k of pow e r.<br />

First, in both e x pe rim e nts the obtaine d patte rn of m e ans w as o p p o site to<br />

the predic tions of the fee ling s-as -inform ation m ode l. He nc e , an inc reas e in<br />

pow e r w ould like ly prov ide strong er e v ide nc e ag ains t rathe r than for the<br />

predic tions of the m ode l, if w e as sum e that additional subje c ts w ould not<br />

reverse the direc tion of the obse rve d e ffe c ts. S e c ond, se v e ral studie s using<br />

c om parable c e ll size s hav e show n that attributional m anipulations, sim ilar<br />

to those w e e m ploy e d, c an unde rm ine the im pac t of subje c tiv e e x pe rie nc e s<br />

like m oods, arousal, or ex pe rience d e ase of rec all (e .g. B ornste in &<br />

D’ Ag ostino, 1 9 9 4 ; S c hw arz & C lore, 1 9 8 3 ; S chw arz et al., 1 9 9 1 ;<br />

S c hw arz , S e rv ay , & Kum pf, 1 9 8 5 ; S inc lair, Mark, & C lore, 1 9 9 4 ; Zanna<br />

& C oope r, 1 9 7 4 ). T hird, the attributional m anipulation use d in Ex pe rim ent<br />

2 w as m ode lle d c lose ly afte r a suc c e ss ful m anipulati on use d by S c hw arz e t<br />

al. (1 9 9 1 ), althoug h it w as targ e te d to address e ase of retrie v al e x pe rie nc e s<br />

in the ir s tudy . Finally , it is w orth ke e ping in m ind that both studie s<br />

atte m pte d to c ounte r a sublim inal e ffe c t w ith e x plic it supralim inal attribution<br />

m anipulati ons.<br />

Im plications for the Affective Prim acy Hypothesis<br />

Tw o of the ke y ® nding s of the c urre nt studie s are c onsiste nt w ith the<br />

affe c tiv e prim ac y hy pothe sis. First, the prim ing effe c t w as im pe rvious to<br />

attributional m anipulations. T his ® nding supports the notion of affe ctc<br />

og nition indepe nde nc e (<strong>Zajonc</strong> , 1 9 8 0 ). This notion is also stre ng the ne d<br />

by the rec e nt e v ide nc e that fac ial e x press ions are proc e ss e d inde pe ndently<br />

of fac ial ide ntity by the e m otional c ircuits of the am y g dala (Adolphs,<br />

Trane l, Dam as io, & Dam asio, 1 9 9 4 ). S e c ond, the prim ing m anipulati on<br />

did not produc e c onsc ious fe e ling s. O ne possible im plicati on is that affe c t<br />

produc ed by sublim inal facial prim es is rudim entary and possibly unc onsc<br />

ious (<strong>Zajonc</strong> , 1 9 9 4 ). S uc h unconsc ious affe c t m ay g uide partic ipants’<br />

judg e m ents and de cisions (se e B ec hara, Dam asio, Trane l, & Dam asio,<br />

1 9 9 7 for a rece nt de m onstration). Another pos sibility is that the pic tures<br />

of e m otional fac ial e x press ions w e re ins uf® c ie nt to trig g e r e m otional<br />

fe e ling s e v e n if the y had be e n e x pose d supralim inally . 3<br />

3 The e x ac t n ature of affe c tiv e re sponse s e li c ite d by sublim in al and supralim inal pre se ntation<br />

of fac ial e x pre ssion s ne e ds to be in v e stig ated further. How e v e r, a re c e nt study found that<br />

sublim inall y pre se nted e x pre ssions e lic it fac ial EM G re sponse sÐ a se nsitiv e in dic ator of an<br />

unde rly in g af fe c tiv e state (De Groot, 1 9 9 6 ; C ac iopp o, Pe tty, Losh, & Kim , 1 9 8 6 ). Furtherm<br />

ore , the pre fe re nc e s e lic ite d by sublim inal fac ial pri m in g c om bin e additiv e ly w ith pre fe re<br />

nc e s e lic ite d by m e re e x posure (M urphy , M onahan, & Zajon c , 1 9 9 5 ). The additiv e pattern<br />

sug g e sts that pre f e re nc e s form e d by bo th pro c e dure s are rather undiff e re ntiate d and in de -<br />

pe nde nt of assoc iated c ontent.


458 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

The re w e re also tw o se ts of ® nding s that w e re not pre dic te d by the<br />

affe c tiv e prim ac y . First, in Ex pe rim e nt 1 judg e m e nts of stim uli in bloc k 2<br />

w e re sig ni® c antly and pos itiv e ly c orrelate d w ith judg e m ents of the sam e<br />

stim uli in bloc k 1 . This oc c urred de spite the fac t that in block 2 the stim uli<br />

w e re paired w ith opposite affe c tiv e prim e s. The fac t that s ubje c ts retaine d a<br />

prev iously form e d e v aluati on s ug g e sts that e v aluati v e c onditioning<br />

oc c urred. Ev aluati v e c onditioning require s a link be tw e e n affe c t and<br />

m e m ory for the stim ulus, and sug g e sts that affe c t inte rac ts w ith cog nition<br />

on a le v e l of prim itiv e operations (Kros nick e t al., 1 9 9 2 ; Le v e y & Martin,<br />

1 9 9 0 ). S e c ond, affe c tiv e prim ing w as se nsitiv e to v arious in¯ ue nc e s. In<br />

both e x perim ents , the prim ing e ffe c t w as in¯ ue nc e d by e x pe rim e ntal<br />

instruc tions. In both e x pe rim e nts, the prim ing w as not obtaine d in the<br />

se c ond bloc k of trials. In Ex pe rim ent 2 the stre ng th of prim ing v arie d<br />

w ith the respons e s to the m usic . The se res ults sug g es t that, contrary to<br />

the strong thes is of inde pe nde nce , the affe c tiv e sy ste m is se nsitiv e to<br />

v arious m odify ing in¯ ue nc e s (se e also Parrott & S chulkin, 1 9 9 3 ). How -<br />

e v e r, as w ill be disc us se d late r, the av ailable ® nding s sug g e st that the<br />

obtaine d e ffe c ts m ay re¯ e c t an affe c t-affe c t interac tion, or an inte rac tion<br />

be tw e e n affe c t and low -lev el c og nition.<br />

The prese nt e x pe rim e nts w ere not de sig ne d to te st a spe ci® c alte rnativ e<br />

to the fe e ling s -as-inform ati on m ode l, re¯ e c ting the lac k of s uf® c ie ntly<br />

spe c i® e d proce ss ac c ounts. How e v e r, it is use ful to e v aluate how othe r<br />

ass um ptions about the proc e sse s unde rly ing affe c tiv e prim ing m ay<br />

ac c ount for the prese nt ® nding s. The re are four diffe rent ty pe s of e x planatory<br />

m ode ls that m ay be use d to ac c ount for the pres e nt ® nding s: (1 )<br />

se m antic ac c ounts; (2 ) atte ntion/pe rce ptual m e m ory acc ounts; (3 )<br />

response stag e prim ing ac c ounts; and (4 ) prefe renc e m isattribution<br />

ac c ounts .<br />

Se m a n tic a c c o u n ts. Thes e ac c ounts, incom pati ble w ith the affe c tiv e<br />

prim ac y , e x plain affe c tiv e prim ing e ffe c ts by m ediati on of proc e sse s<br />

c onc erne d w ith m e aning . A purely se m antic m ode l assum es that the re<br />

is no diffe renc e be tw e e n prim ing w ith affe c tiv e and nonaffe c tiv e stim uli.<br />

In both c ase s, prim ing inv olv e s proc e ss ing of the m e aning of the prim e<br />

and inte rpretati on of the targ e t in term s of ac tiv ate d c onc e pts (e .g.<br />

Hig g ins, Rhole s, & Jone s, 1 9 7 7 ). Re ce nt e v ide nce arg ue s ag ainst a<br />

purely se m antic m ode l. In e x pe rim e nts by Murphy and Z ajonc (1 9 9 3 )<br />

prim ing at v e ry low e x posure c onditions w as obtai ne d only w ith affe c tiv e<br />

prim es . Nonaffe ctiv e prim e s produce d prim ing only w he n e x posure tim e s<br />

w e re inc rease d, resulting in s upralim inal e x posure. This sug g e sts that<br />

affe c tiv e prim ing , at least unde r sublim inal c onditions, inv olv e s som e<br />

unique proc e s se s.<br />

The log ic of an interm e diate affe c tiv e -c og nitiv e m ode l (e.g . B ow e r,<br />

1 9 9 1 ; Forg as, 1 9 9 2 ) allow s for a prim ary affe c tiv e respons e to be e lic ite d


SU B LIM IN A L A FFECT IV E P RIM IN G 459<br />

by a prim e. This aff ec tiv e response , how e v e r, le ads to affe c t-c ong rue nt<br />

judg e m ents o n ly if it rende rs affe c t-c ong rue nt c onc e pts ac c e ssible . Unfortunate<br />

ly , the s em antic ne tw ork m ode l has trouble e x plaining s ev e ral affe c -<br />

tiv e in¯ uenc e phe nom ena (se e disc uss ions by Nie de nthal et al., 1 9 9 4 ;<br />

Forg as, 1 9 9 5 ; B arg h, C haike n, Ray m ond, & Hy m e s, 1 9 9 6 ).<br />

He nce , s ati sfac tory ex planations of the affe c tiv e prim ing e ffe c ts are<br />

like ly to be offe red by m ode ls that assig n an im portant role to the early<br />

affe c t and do not as sum e hig h-le v e l se m antic m e diation. Three diff erent<br />

ac c ounts sati sfy this c rite rion. The y postulate e arly affe c tiv e s tate s and<br />

assum e that the m e diati ng proc e ss is low -le v e l and autom atic . The se<br />

ac c ounts diffe r in loc ating the proc e ss responsible for prim ing at diff erent<br />

stag e sÐ atte ntion, pe rce ption, or res pons e g e ne ration.<br />

Atte n tio n a n d p e rc e p tu a l m e m o ry a c c o u n ts. O ne w ay in w hic h affe c t c an<br />

in¯ ue nce judg e m ents is by c hang ing the pe rce ption of targ e ts v ia a fronte<br />

nd prim ing m ec hanis m . This m echanis m m ay inv olv e c hang e s in atte ntion<br />

to affe c t-cong rue nt s tim uli (B roadbent & B roadbe nt, 1 9 8 8 ; De rrybe rry &<br />

Tuc ke r, 1 9 9 4 ). A lte rnati v e ly , the m e c hanism c ould inv olv e prim ing of<br />

affe c t-c ong rue nt perce ptual represe ntati ons (Nie de nthal e t al., 1 9 9 4 ).<br />

From both the atte ntional and perce ptual m e m ory pe rspe c tiv e , the affe c -<br />

tiv e prim ing e ffe c ts m ay be ex plaine d by ass um ing that e x posure to happy<br />

prim e s inc rease s e ithe r atte ntion to, or ac tiv ation of, positiv e aspe c ts of the<br />

targ e t ide og raph, w he reas e x pos ure to angry prim e s inc rease s e ither atte ntion<br />

to, or ac tiv ati on of, its ne g ativ e aspe c ts. C onsiste nt w ith this assum ption,<br />

w e found in E x pe rim e nt 2 that s ublim inal prim e s w e re m ost e ffe c tiv e<br />

w he n the affe ctiv e v ale nc e of the prim e s w as c ong ruent w ith the affe c t<br />

m e ntione d in the e x pe c tation instruc tions. T hat is , happy prim es tende d to<br />

be m ost e ffe c tiv e w he n the e x pe ctati on m anipulation le d subjec ts to e x pe c t<br />

positiv e af fe c t, and ne g ative prim es w e re m ost effe c tiv e w he n the e x pe c tation<br />

m anipulati on le d subje c ts to e x pe c t ne g ative e ffec t. This m ay e ithe r<br />

indic ate that the e x pe c tati on m anipulation its elf in¯ ue nc e d subje c ts ’ affe c t,<br />

resulting in diffe rential atte ntion to, or ac tiv ati on of, affe c t-c ong rue nt<br />

m ate rial, or that subje c ts ’ e x pe c tati ons in¯ ue nce d pe rce ption inde pe nde nt<br />

of an in¯ ue nc e on the ir affe c t. 4<br />

Com plic ati ng this is sue , how e v e r, w e did not obse rve a parallel effe c t in<br />

Ex pe rim ent 1 , w he re the im pact of the prim es w as only e nhance d by the<br />

4 In g e ne ral, the plausibili ty of the pe rc e ptual/attentional e x planation for the c urre nt<br />

studie s re sts on the tw o assum ptions. Firs t ide og raphs as targ e ts hav e fe atures that c an be<br />

diffe re ntiall y attende d to, ac tiv ated, or interpre ted de pe n din g on the prim e . S e c ond, suc h<br />

diffe re ntiall y ac tiv ated or in terpre ted fe ature s c an v ary in an affe c tiv e v alue . B oth assum ptions<br />

are c o nsis tent w ith ® nding s f rom e x pe rim e ntal ae sthetic s that show c ontex tual de pe n-<br />

de nc e of pe rc e ptio ns, in terpre tations, and pre fe re nc e s for sim ple v is ual e le m e nts (B e rly ne ,<br />

1 9 7 4 ; M c Manus, 1 9 8 0 ).


460 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

e x pe ctati on m anipulation w he n s ubje cts w e re inform e d that the y w ould be<br />

e x pose d to sublim inal prim e s, but w e re not inform e d about how thes e<br />

prim es m ig ht affec t the m . Althoug h this latte r ® nding m ay als o be trace d<br />

to diffe renc e s in atte ntion to the tas k, the se diffe rence s w e re inde pe nde nt of<br />

v ale nc e , in c ontrast to the ® nding s in Ex perim ent 2 . A nothe r ® nding that<br />

m ay be relate d to nonspe c i® c diffe re nc e s in atte ntion is the disappe aranc e<br />

of prim ing in the se c ond bloc k of both e x pe rim e nts .<br />

Re sp o n se sta g e p rim in g a c c o u n ts. In 1 9 8 4 , <strong>Zajonc</strong> and Markus propose d<br />

that affe c t c an inte rac t w ith c og nition be c ause both sy ste m s hav e direc t<br />

links to the m otor respons e s y ste m . A c c ording to this ide a, low -le v e l<br />

affe c tiv e inform ation in¯ ue nc e s judg e m ent by biasing the se le c tion of a<br />

m otor respons e , w hic h is norm ally unde r the c ontrol of the c og nitiv e<br />

sy ste m . Inte resting ly , rec e nt rese arc h s ug g e s ts that se nsory inform ation<br />

m ay inde ed s pe c ify response param e te rs w hile by pass ing hig he r le v e l<br />

represe ntati on. For e x am ple , Ne um ann and Klotz (1 9 9 4 ) found that a<br />

sublim inal, nonaffe c tiv e v isual prim e m ay direc tly trig g e r a m otor<br />

response . Othe r ® nding s sug g e st that affe c tiv e stim uli m ay hav e e spe cially<br />

pow e rful unm ediate d effe c ts on the re spons e sy ste m (B arg h e t al., 1 9 9 6 ;<br />

C ac ioppo, C rite s, Gardne r, & B e rntson, 1 9 9 4 ; C acioppo, Prie ste r, &<br />

B e rnts on, 1 9 9 3 ; Lang , 1 9 9 5 ; O hm an & S oare s, 1 9 9 4 ). A lthoug h prov oc ativ<br />

e , the response stag e prim ing ac c ount c annot e x plain the instruc tion<br />

e ffe c ts and the disappe aranc e of prim ing e ffe c ts in the se c ond bloc k of<br />

both e x pe rim e nts .<br />

Pre fe re n c e a ttrib u tio n a c c o u n ts. Finally , a m odi® ed attributional m ode l<br />

c an be propose d. Ac c ording to this m ode l, sublim inal affe c tiv e prim e s<br />

autom atically ac tiv ate an assoc iate d e v aluation (Faz io e t al., 1 9 8 6 ; B arg h<br />

e t al., 1 9 9 2 ), w hic h is m isinte rpreted as an e v aluati v e response to the<br />

supralim inally prese nte d targ e t. Althoug h this inte rpretati on s hare s the<br />

attributional as sum ption of the fe e ling s-as-inform ation m ode l, it doe s not<br />

postulate a role for e x pe rie nc e d ``fe e ling s’ ’ that are ac c e ssible for c onsc<br />

ious inferenc e s. He nc e , (m is )attribution m anipulations w ould not be<br />

e x pe cte d to in¯ ue nc e s ubje c ts’ autom ati c prefe renc e judg e m e nts. Moreov<br />

er, this ac c ount doe s not ne e d to trac e the e lim ination of affe ctiv e prim ing<br />

e ffe c ts to any aw are ne ss of an irrele v ant source that unde rm ine s the inform<br />

ational v alue of subje c tiv e e x pe rie nc e s. T his w ay , the m ode l c an ac c ount<br />

for the bloc k e ffe c ts obse rve d in E x pe rim e nt 1 . From this pe rspe c tiv e ,<br />

subjec ts m ay hav e ``m isread’ ’ the ir prefe renc e response to the sublim inal<br />

prim es as a response to the ideog raphs prese nte d in the ® rst bloc k, presum -<br />

ably be c ause the ne utral ide og raphs did not e lic it any sy ste m ati c preference s<br />

on the ir ow n. How e v e r, w he n the sam e ide og raphs w e re show n in the se c ond<br />

bloc k, the y had already ac quired an e v aluation during the prev ious e nc ounte<br />

r. Conse que ntly , the ide og raphs c ould e lic it the ir ``ow n’ ’ prefe renc e<br />

response , resulting in a lac k of im pac t of the prefe renc e e lic ited by the prim e .


SU B LIM IN A L A FFECT IV E P RIM IN G 461<br />

In sum , the e m e rge nce of affe c tiv e prim ing e ffe cts, the lac k of disc ounting<br />

and aug m e ntati on e ffe c ts, the abs e nc e of fee ling s, and the bloc k e ffe cts<br />

obse rv e d in Ex perim ent 1 are c om patible w ith the preference m isattribution<br />

ac c ount. How e v e r, this ac c ount prov ide s no c om pelling e x planati on<br />

for the instruc tional e ffe c ts obse rve d in both e x pe rim e nts and disappe aranc<br />

e of prim ing in the se c ond bloc k of E x pe rim e nt 2 .<br />

Im plications for the Feeling-as-inform ation M odel<br />

Re se arc h on affe c tiv e in¯ ue nc e s on judg e m e nt has predom inantly<br />

address ed the im pac t of c onsc ious ly e x pe rie nc e d affe c t in the form of<br />

m oods and e m otions. Most of this rese arc h is c onsiste nt w ith the assum ption<br />

that affe ctiv e in¯ ue nc e s are m e diate d by the use of one’ s affe c tiv e<br />

state s as a source of inform ation, as c onc e ptualise d in the fe e ling s-asinform<br />

ati on m ode l (se e C lore e t al., 1 9 9 4 ; S chw arz & C lore, 1 9 9 6 for<br />

rec ent rev ie w s). T he alte rnati v e ac c ounts disc us se d e arlie r, the se m antic<br />

ne tw ork acc ount, the re sponse prim ing ac count, and the atte ntion and<br />

pe rce ptual m em ory acc ount, do not e x plain the repe ate dly obse rve d de pe nde<br />

nc e of affe c tiv e in¯ ue nc e s on the pe rce iv ed inform ational v alue of<br />

indiv iduals ’ fe e ling s (S c hw arz & C lore, 1 9 9 6 ).<br />

The prese nt ® nding s, how e v e r, raise the possibility that s om e affe c tiv e<br />

in¯ ue nce s are unlike ly to follow the infe rential patte rns obse rve d for<br />

m oods and e m otions (s ee Forg as, 1 9 9 5 , p. 4 3 for a disc ussion of this<br />

issue ). T his m ay be the c as e w he n affe c tiv e reactions fee d into currently<br />

unde rspe c i® e d autom ati c proc e sse s, w hic h result in a c hang e in judg e -<br />

m e nts. This m ay also be the c as e w he n affe c tiv e reac tions are not represe<br />

nte d in a form of c onsc ious fe e ling . He nc e , the prese nt ® nding s do not<br />

fals ify the fe e ling s-as-inform ation m ode l as an ac c ount for the im pac t of<br />

subje c tiv e e x pe rie nc e s. Rathe r, the y sug g e st that the log ic unde rly ing the<br />

m ode l m ay be inapplic able to c onditions w he re the affe c t induc tion doe s<br />

not result in c ons c iously e x pe rie nc e d fe e ling s.<br />

Moreov e r, as indic ate d e arlie r, the prese nt ® nding s m ay be inte rpreted as<br />

sug g e sting that affe c tiv e prim ing inv olv e s a (m is )attribution of prefe renc e<br />

response s instead of a m isattribution of fe e ling s . A te st of this possibility<br />

w ould require that the e x pe rim e ntal instruc tions targ e t (m is)attribution of<br />

prefe renc e response s rathe r than of fe e ling s and ``g ut reac tions’ ’ . Note ,<br />

how e v e r, that s uc h disc ounting predic tions ag ain ass um e that the e v aluativ<br />

e response and a targ et stim ulus are linke d by an infe rential ``attributional’<br />

’ proc e ss, and not by a m ore prim itiv e proc e ss. Furthe rm ore, the<br />

disc ounting predic tion as sum es that subje c ts are able to de te c t and s e parate<br />

tw o liking response sÐ the ® rst to the prim e , and the se c ond to the ide og<br />

raph. If the liking response s w e re not se parable for som e reason (e .g .<br />

be c ause s ubje cts e x pe rie nce only one sing le , or one c om bine d, prefe renc e


462 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

response that is sim ultane ous to the pe rce ption of the ide og raph), attributional<br />

m anipulati ons w ould be ine f® c ie nt. Ne v e rthe le ss, the obtai ne d<br />

affe c tiv e prim ing effe c ts c ould still re¯ e c t a m isreading of the prefe renc e<br />

response to the prim e as a prefe renc e response to the ide og raph. 5<br />

In c onc lusion, the pre se nt ® nding s sug g e s t that the re are g ood theoretic<br />

al and e m piric al reas ons to c onside r m ode ls of affe c tiv e in¯ ue nce s on<br />

e v aluativ e judg e m ents that do not require a m e diati ng role of c ons ciously<br />

e x pe rie nc e d fe e ling s. The ® nding s also sug g e st that som e ty pes of affe c tiv e<br />

in¯ ue nc e s m ay not be respons iv e to attributional m anipulations des ig ne d to<br />

v ary the ir inform ati onal v alue . A lthoug h our present unde rstanding of the<br />

prec ise m e c hanism s unde rly ing s ublim inal affe c tiv e prim ing rem ains tentativ<br />

e , current rese arc h sug g e sts that the af fe c tiv e prim ac y hy pothe s is<br />

prov ide s a fram e w ork that m ay g uide the se arc h for a the oretic al ac c ount<br />

of this bas ic phe nom e non.<br />

M anusc ript re c e iv e d 3 1 O c tobe r 1 9 9 5<br />

Re v is e d m anusc rip t re c e iv e d 9 De c e m be r 1 9 9 6<br />

REFER EN C ES<br />

A dolphs, R., T rane l, D., Dam asio, H., & Dam asio, A. ( 1 9 9 4 ). Im paire d re c og nitio n of<br />

e m o tio n in fac ial e x p re ss ion s f ollow ing b ilate ral dam ag e to the hu m an am y g dala.<br />

Na tu re , 3 7 2 , 6 6 9 ± 6 7 2 .<br />

A g g le ton, J.P., & M is hkin, M . (1 9 8 6 ). The am y g dala: S e nsory g atew ay to the e m otions. In<br />

R. Plutchik & H. Ke lle rm an ( Eds.), Em o tio n : Th e o ry, re se a rc h , a n d e xp e rie n c e (V ol. 3 ,<br />

pp. 2 8 1 ± 2 9 9 ). O rlando, FL : A c ade m ic Pre ss.<br />

B arg h, J.A . (1 9 8 9 ). C onditio nal autom atic ity : V arie tie s of autom atic in¯ ue n c e in so c ial<br />

pe rc e ption an d c og nition. In J.S . Ulem an & J.A . B arg h (Eds.), U n in ten d e d th o u g h t<br />

(pp. 3 ± 5 1 ). Ne w York: G uilford Pre ss.<br />

B arg h, J.A ., Chaike n, S ., Gov e nde r, R., & Pratto, F. ( 1 9 9 2 ). The g e ne rali ty of the autom atic<br />

attitude ac tiv atio n e ffe c t. Jo u rna l o f Pe rso n a li ty a n d So c ia l Psyc h o lo g y, 6 2 , 8 9 3 ± 9 1 2 .<br />

5 The is sue of se parabili ty and dis c rim inabil ity of affe c tiv e re sponse s is o f k e y im portanc e<br />

he re . This is furthe r hig hli g hted by re c e nt ® ndin g s by Gaspe r and C lo re (1 9 9 6 ). The se<br />

authors in form e d subje c ts abo ut plausib le situational c ause s of the ir m oods, but obtaine d<br />

the usual disc ounting e ffe c t only f or subje c ts hig h on attention to e m otio n. In c ontrast,<br />

subje c ts low on attention to e m otio n show e d an inc re ase in m ood-c ong rue nt judg e m e nt unde r<br />

disc ounting c onditions, pre sum ably re ¯ e c ting e nhanc e d attention to the ir fe e ling s. This type<br />

of e v id e nc e sug g e sts that w he n de tectabili ty of affe c tiv e c hang e s is situationally or dis po-<br />

sitionally low , a dis c ountin g m anipulation m ay ac tually e nhanc e affe c t-c ong rue nt e f fe c ts.<br />

The ® nding s of Gaspe r and C lo re (1 9 9 6 ) are c on sis tent w ith the patte rn obse rv e d in<br />

Ex pe ri m e nt 1 , w he re fore w arning le d to an inc re ase in the affe c tiv e prim ing e ffe c t. How -<br />

e v e r, in Ex pe rim e nt 1 fore w arne d subjects did not re port e x pe rie nc in g fe e lin g s, nor did the y<br />

re port re ly ing m ore on ``g ut re ac tions’ ’ w hil e m aking judg e m e nts of ide og raphs, thus<br />

sug g e stin g that fore w arning in c re ase d attention to the task rather than attention to e m otion.


B arg h, J.A ., C haike n, S ., Ray m ond, P., & Hy m e s, C . (1 9 9 6 ). The autom atic e v aluation<br />

e ffe c t: Unc onditio nal autom atic attitude ac tiv atio n w ith a pronunc iation task . Jo u rn a l o f<br />

Exp e rim e n ta l So c ia l Psyc h o lo g y, 3 2 , 1 0 4± 1 2 8 .<br />

B arg h , J.A ., & P ie trom o nac o, P . (1 9 8 2 ). A u tom atic in form ation p roc e s sin g an d soc ial<br />

pe rc e ption: The in¯ ue nc e of trait inform atio n pre se nted outside of c onsc io us aw are ne ss<br />

on im pre ssio n form ation. Jo u rn a l o f Pe rso n a lity a n d Soc ia l Psyc h o lo g y, 4 3 , 4 37 ± 4 4 9.<br />

B e c hara, A ., Dam asio, H., Trane l, D., & Dam asio, A .R. (1 9 9 7 ). De c iding adv antag e ously<br />

be fore know in g the adv antag e ous strategy . Sc ience, 2 7 5 , 1 2 9 3 ± 1 2 9 5 .<br />

B e rly ne , D.E . (1 9 7 4 ). Stud ie s in th e n e w e xp e rim e n ta l a e sth e tic s: step s to w a rd a n o b je c tiv e<br />

p syc h o lo g y o f a e sthe tic a p p re c ia tio n . W ashing ton, DC : He m is phe re .<br />

B ornstein, R.F., & D’ A g ostin o, P.R. (1 9 9 4 ). The attribution and dis c ounting of pe rc e p tual<br />

¯ ue nc y : Pre li m inary tests o f a pe rc e ptual ¯ ue nc y /attributional m ode l of the m e re e x -<br />

posure e ffe c t. Soc ia l C o g n itio n , 1 2 , 1 0 3± 1 2 8 .<br />

B ow e r, G.H . (1 9 9 1 ). Mood c ong ruity of soc ial judgm e nts. In J.P. Forg as (Ed.), Emo tio n a n d<br />

so c ia l ju d g m en ts (pp. 3 1 ± 5 4 ). O xford, UK: Pe rg am on.<br />

B ro adb e n t, D., & B ro adb e nt, M . ( 1 9 8 8 ). A n x ie ty and atten tional bias: S tate and trait.<br />

C og n itio n a n d Em o tio n , 2 , 1 6 5 ± 1 8 3 .<br />

C ac ioppo, J.T ., C ri tes, S .L ., Gardne r, W .L ., & B e rn tson, G.G. (1 9 9 4 ). B ioe le c tric al e c hoe s<br />

from e valuativ e c ateg oriz ations. I: A late positiv e brain potential that v arie s as a func tion of<br />

trait ne g ativ ity and e x tre m ity . Jo u rn a l o f Pe rso n a lity a nd So c ia l Psyc h o lo g y, 6 7 , 1 1 5 ± 1 2 5 .<br />

C ac ioppo, J.T ., Pe tty, R.E ., Losh, M .E ., & Kim , H.S . ( 1 9 8 6 ). Electrom y og raphic ac tiv ity<br />

ov e r fac ial m usc le re g ion c an dif fe re ntiate the v ale nc e and intensity of affe c tiv e re ac -<br />

tions. Jo u rn a l o f Pe rso n a lity a n d So c ia l Psyc h o lo g y, 5 0 , 2 6 0 ± 2 6 8 .<br />

C ac iop po , J.T., P ri e s ter, J.R., & B e rntso n, G .G. ( 1 9 9 3 ). Rud im e ntary de te rm in an ts of<br />

attitude s: II. A rm ¯ e x ion and e x tensio n hav e diff e re ntial e ffe c ts on attitude s. Jo u rn a l<br />

o f Pe rso n a li ty a n d So c ia l Psyc h o lo g y, 6 5 , 5 ± 1 7 .<br />

C lore , G.L ., S c hw arz , N., & C onw ay , M . (1 9 9 4 ). A ffe c tiv e c ause s and c onse que nc e s of<br />

soc ial inf orm ation proc e ssi ng . In R.S . W y e r & T.K. S rull ( Eds.), Han d b o ok o f so c ia l<br />

c o g n itio n (2 nd e d.), (V ol. 1 , pp. 3 2 3 ± 4 1 7 ). Hil ls dale , NJ: Law re nc e Erlbaum Assoc iates<br />

Inc .<br />

Dam asio , A .R. (1 9 9 4 ). D e sc a rtes’ e rro r: E mo tio n , re a so n a n d the h u m a n b ra in . Ne w Y ork:<br />

Grosse t/Putnam .<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 463<br />

De Groot, P. (1 9 9 6 ). F a c ia l EMG a n d n o n c o n sc io u s a ffectiv e p ri m in g . Unpubli she d M asters<br />

thesis, Univ e rsity of A m sterdam .<br />

De rry be rry , D., & T uc ke r, D.M . (1 9 9 4 ). M otiv atin g the foc us of attention. In P.M . Nie -<br />

de nthal & S . K itay am a ( Eds.), Th e h e a rt’ s e ye : Em otio n a l in ¯ u e n c e s in p e rc e p tio n a n d<br />

a tten tio n (pp. 1 6 7 ± 1 9 6 ). S an Die g o, C A : A c ade m ic Pre ss.<br />

Edw ards, K . (1 9 9 0 ). The interplay of affe c t and c og nition in attitude form atio n and c hang e .<br />

Jo u rn a l o f Pe rso n a lity a n d Soc ia l Psyc h o lo g y, 5 9 , 2 02 ± 2 1 6.<br />

Faz io, R.H ., S anbonm atsu, D.M ., Pow e ll , M.C ., & Karde s, F.R. (1 9 8 6 ). O n the autom atic<br />

ac tiv ation of attitude s. Jo u rn a l o f Pe rso na lity a n d So c ia l Psyc h o lo g y, 5 0 , 2 2 9 ± 2 3 8 .<br />

Fodor, J.A . (1 9 8 3 ). Th e m o d u la rity o f min d . C am bridg e , M A: M IT Pre ss.<br />

Forg as, J.P. (1 9 9 2 ). A ffe c t in soc ial judg m e nts and de c is ions: A m ulti- proc e ss m ode l. In<br />

M.P. Zanna (Ed.), Ad v a n c e s in e xp e rim e n tal so c ia l p syc h o lo g y (V ol. 2 5 , pp. 2 2 7 ± 2 7 5 ).<br />

Ne w Y ork: A c ade m ic Pre ss.<br />

Forg as, J.P. ( 1 9 95 ) . M ood and ju dg m e nt: The affe c t in fusion m o de l (AIM ). Psyc h o lo g ic a l<br />

Bu lle tin , 1 1 7 , 3 9 ± 6 6 .<br />

Gas pe r, K., & C lore , G.L . (1 9 9 6 ). Mo o d a n d ju d g m e n t: Th e in ¯ u e n c e o f a tten tio n to<br />

e m o tio n . Pape r pre se nted at the 6 8 th A n nual M e e ting of M idw e stern Ps y c h olog ical<br />

Assoc iatio n, C hic ag o.


464 W IN K IELM A N , Z AJO N C, S CH W A RZ<br />

Hig g ins , E.T., Rhole s, W .S ., & Jone s, C .R. (1 9 7 7 ). C ateg ory ac c e ssibil ity and im pre ssion<br />

form atio n. Jo u rn a l o f Exp e rim e n tal So c ia l Psyc h o lo g y, 1 3 , 1 4 1 ± 1 5 4 .<br />

Ke ll e y , H.H. ( 1 9 7 2 ). C au sal sc he m ata and the attributio n p roc e ss. In E.E . Jone s , D.E .<br />

Kanouse , H.H. Ke lle y, R.E . Nis be tt, & S . V alins (Eds.), Attrib u tio n : Pe rc e ivin g th e<br />

c a u se s o f b e h a v io r. M orris tow n, NJ: Ge ne ral Le arning Pre ss.<br />

Kitay am a, S . (1 9 9 1 ). Im pairm e nt of pe rc e ptio n by positiv e and ne g ativ e affe c t. C o g n itio n<br />

a n d Em otio n , 5 , 2 5 5 ± 2 7 4 .<br />

Krosnic k, J.A ., B e tz, A .L ., Jussim , L.J., & Ly nn, A .R. (1 9 9 2 ). S ubli m in al c onditio ning of<br />

attitude s. Pe r son a lity a n d So c ia l Psyc h o lo g y Bu lle tin , 1 8 , 1 5 2 ± 1 6 2 .<br />

Kunst-W ils on, W .R., & <strong>Zajonc</strong> , R.B . (1 9 8 0 ). A ffe c tiv e dis c rim inatio n of stim uli that c annot<br />

be re c og niz e d. Sc ie n c e , 2 0 7 , 5 5 7 ± 5 5 8 .<br />

Ladavas, E., C im atti, D., De lP e s c e , M ., & Tuoz z i, G. (1 9 9 3 ). Em otio nal e v aluation w ith<br />

and w ithou t c o nsc iou s stim ulus identi® c ation: E v ide nc e from a s plit-brain patie n t.<br />

C o g n itio n a n d Em o tio n, 7 , 9 5 ± 1 14 .<br />

L an g , P .J. (1 9 9 5 ). The e m otion probe : S tudies of m otiv atio n and atte ntio n. Am e rica n<br />

Psyc h o lo g is t, 5 0 , 3 7 2 ± 3 8 5 .<br />

Le Doux , J.E . (1 9 8 9 ). C ogn itiv e -e m otional interac tio ns in the brain . C o gn itio n a n d Em o tio n ,<br />

3 , 2 6 7 ± 2 8 9 .<br />

Le Doux , J.E. (1 9 9 3 ). Em otional ne tw orks in the brain . In M. Le w is & J.M. H av il and (Eds.),<br />

Ha n d b o o k o f e m o tio n s (pp. 1 0 9 ± 1 1 8 ). Ne w Y ork: G uilford Pre ss.<br />

Le v e y, A .B ., & Martin, I. (1 9 9 0 ). Ev aluativ e c onditio nin g : O ve rv ie w and f urther options.<br />

C o g n itio n a n d Em o tio n, 4 , 3 1 ± 3 7.<br />

M c M anus, I.C . (1 9 8 0 ) . The ae sthe tic s of sim ple ® g ure s. Br itis h Jo u rn a l o f Psyc ho lo g y, 7 1 ,<br />

5 0 5 ± 5 2 4 .<br />

M urphy , S .T ., M onahan, J.L ., & Zajo nc , R.B . (1 9 9 5 ). Additiv ity of nonc onsc io us af fe c t:<br />

C om bin e d e f fe c ts of prim ing and e x posure . Jo u rn a l of Pe rso n a li ty a n d So c ia l Psyc h o lo<br />

g y, 6 9 , 5 8 9 ± 6 0 2 .<br />

M urphy , S .T ., & <strong>Zajonc</strong> , R.B . (1 9 9 3 ). A ffe c t, c og nition, and aw are ne ss: A ffe c tiv e prim ing<br />

w ith optim al and suboptim al stim ulus e x pos ure s . Jo u rn a l o f Pe rso n a lity a n d So c ia l<br />

Psyc h o lo g y, 6 4 , 7 2 3 ± 7 3 9 .<br />

Ne um ann, O ., & Klo tz, W . ( 19 9 4 ). M otor re spo nse s to no nre po rtable , m aske d stim uli:<br />

W he re is the lim it of dire c t param e ter spe c i® c ation? In M . M osc ov itc h & C . Um ilta<br />

(Eds.), Atten tio n a n d p e rfo rma n c e . XV: C o n sc io u s a n d n o n c o n sc io u s in forma tio n p ro c e ssin<br />

g . C am bridg e , M A: M IT Pre ss.<br />

Nie de n thal, P.M . (1 9 9 0 ). Im plic it pe rc e ptio n of affe c tiv e in form ation. Jo u rn al o f Exp e rim<br />

e n ta l So c ia l Psyc h o lo g y, 2 6 , 5 0 5 ± 5 2 7 .<br />

Nie de n thal, P.M ., S e tterlund, M .B ., & Jone s, D.E . (1 9 9 4 ) . Em otio nal org aniz ation of pe rc<br />

e ptual m e m ory . In P.M . Nie de nthal & S . Kitay am a (E ds.), Th e h e a rt’ s e ye . Em o tio n a l<br />

in ¯ u e n c e s in p e rc e p tio n a n d a tten tio n (pp. 8 7 ± 1 1 3 ). S an Die g o, C A : A c ade m ic Pre ss.<br />

O hm an, A ., & S oare s, J.J.F. ( 1 99 4 ) . ``Unc onsc io us anx ie ty ’’ : Phobic re sponse s to m aske d<br />

stim uli. Jo u rn a l o f Ab n o rm a l Psyc h o lo g y, 1 03 , 2 3 1 ± 24 0 .<br />

Parrott, W .G., & S c hulk in , J. ( 1 9 9 3 ). Ne uropsy c holog y and the c og nitiv e nature of the<br />

e m otions. Co g n itio n a n d Em o tio n , 7 , 4 3 ± 5 9 .<br />

Pratto, F. (1 99 4 ). Consc io usne ss and autom atic e v alu atio n. In P.M . Nie de nthal & S . Ki-<br />

tayam a (Eds.), Th e h e a rt’ s e ye . Em o tio n a l in ¯ u e n c e s in p e rc e p tio n a n d a tten tio n ( pp. 6 7 ±<br />

8 5 ). S an D ie g o, C A: A cade m ic Pre ss.<br />

Pratto, F., & John, O .P . (1 9 9 1 ). Autom atic v ig il anc e : The autom atic g rabbin g pow e r of<br />

ne g ativ e soc ial inform atio n. Jo u rn a l o f Pe rso n a li ty a n d So c ia l Psyc h o lo g y, 6 1 , 3 8 0 ± 3 9 1 .<br />

Ross, M., & O lson, J.M . (1 9 8 1 ). An e x pe c tanc y -attributio n m ode l of the e ffe c ts of plac e -<br />

bos. Psyc h o lo g ical R ev ie w , 8 8 , 4 0 8 ± 4 3 7 .<br />

S c hw arz , N. (1 9 9 0 ). Fe e li ng as in form ation: In form ational and m otiv ational func tions of


affe c tiv e states. In E .T . Hig g ins & R.M . S orre ntin o (Eds.), Ha n db o o k o f m otiv atio n a n d<br />

c o g n itio n (pp. 5 2 7 ± 56 1 ). Ne w Y ork: Guil ford Pre ss.<br />

S c hw arz , N ., B le s s , H., S trac k , F., K lum p p , G ., Ritte n au e r- S c ha tka, H., & S im o n s ,<br />

A. (1 9 9 1 ) . Ease of re trie v al as in form atio n: A nother look at the av ail abili ty he uris tic .<br />

Jo u rn a l o f Pe rso n a lity a n d Soc ia l Psyc h o lo g y, 6 1 , 1 95 ± 2 0 2.<br />

S c hw arz , N., & C lore , G.L . (1 9 8 3 ). Mood, m is attrib utio n, and judg m e nts of w e ll -be in g :<br />

Inform ativ e and dire c tiv e f unc tions of af fe c tiv e state s. Jo u rn a l o f Pe rso n ali ty a n d So c ia l<br />

Psyc h o lo g y, 4 5 , 5 1 3 ± 5 2 3 .<br />

S c h w arz , N., & C lore , G.L . (1 9 8 8 ). How do I f e e l abo ut it? Inform ativ e fu nc tio ns of<br />

affe c tiv e states. In K. Fie dle r & J. Forg as (Eds.), Affec t, c o g n itio n , a n d so c ia l b e h a v io r<br />

(pp. 4 4 ± 6 2 ). Toronto, C anada: Hog re f e .<br />

S c hw arz , N., & C lo re , G.L . (1 9 9 6 ). Fe e li ng s and phe nom e nal e x pe rie nc e s. In E.T . Hig g ins<br />

& A . K rug lanski (Eds.), So c ia l p syc h o lo g y: A h an d b o ok o f b a sic p rin c ip le s (pp. 4 3 3 ±<br />

4 6 5 ). Ne w Y ork: Guilf ord Pre ss.<br />

S c hw arz , N., S e rv ay , W ., & Kum pf, M . (1 9 8 5 ). A ttri butio n of arousal as a m e diator of the<br />

e ffe c tiv e ne ss of fe ar-arousing c om m unications. Jo u rn a l o f Ap p li e d So c ia l Psyc h o lo g y,<br />

1 5 , 7 4 ± 8 4 .<br />

S in c lair, R.C ., Mark, M .M ., & C lore , G.L . (1 9 9 4 ). Mood-re late d pe rsuas io n de pe nds on<br />

(m is )attrib utions. So c ia l C o g n itio n , 1 2 , 3 0 9± 3 2 6 .<br />

S trac k, F. (1 9 9 2 ). The diff e re nt routes to soc ial ju dgm e nts: Ex pe ri e ntial v e rsus in form ational<br />

strateg ie s . In L.L . M artin & A. Te sse r (Eds.), Th e c o n stru c tio n o f so c ia l ju d g m e n ts<br />

(pp. 2 4 9 ± 2 7 6 ). Hill sdale, NJ: L aw re nc e Erlb aum A ssoc iates Inc .<br />

S trac k, F., S c hw arz , N ., B less, H., KuÈ ble r, A ., & W aÈ nk e , M. (1 9 9 3 ). A w are ne ss of the<br />

in ¯ ue nc e as a de term in ant of assim ilation v e rsus c ontrast. Eu ro p e a n Jo u rn a l o f So c ia l<br />

Psyc h o lo g y, 2 3 , 5 3 ± 6 2 .<br />

W e g e ne r, D.T ., & Pe tty, R.E . ( 1 9 95 ) . Fl e x ible c orre c tio n proc e sse s in soc ial judg m e nt: The<br />

role of naiv e the orie s in c orre c tion s for p e rc e ive d bias . Jo u rn a l o f P e rso n a li ty a n d So c ia l<br />

Psyc h o lo g y, 6 8 , 3 6 ± 5 1 .<br />

SU B LIM IN A L A FFECT IV E P RIM IN G 465<br />

Zajo nc , R.B . (1 9 8 0 ) . Fe e ling and thin king : P re fe re nc e s ne e d n o inf e re n c e s . Am e ric a n<br />

Psyc h o lo g ist, 3 5 , 1 1 7 ± 1 2 3 .<br />

<strong>Zajonc</strong> , R.B . (1 9 8 4 ). O n the prim ac y of aff e c t. Am e ric a n Psyc h o lo g is t, 3 7 , 1 1 7 ± 1 2 3 .<br />

<strong>Zajonc</strong> , R.B . (1 9 9 4 ). C an e m otio ns be unc onsc ious? In P. Ekm an & R.J. Dav idson (Eds.),<br />

The n a tu re o f e m o tio n : fun d am e n ta l q u e stio n s (pp. 2 9 3 ± 2 9 7 ). Ne w Y ork: O x ford Univ<br />

e rsity Pre ss.<br />

<strong>Zajonc</strong> , R.B ., & M arkus, H. (1 9 8 4 ). A ffe c t and c og nitio n: The hard in terfac e . In C . Iz ard, J.<br />

Kag an, & R.B . <strong>Zajonc</strong> (Eds.), Em o tio n s, c o g n itio n a n d b e h a v io r (pp. 7 3 ± 1 0 2 ). C am -<br />

bridg e : C am bri dg e Univ e rsity Pre ss.<br />

Zanna, M .P ., & Coope r, J. (1 9 7 4 ). Dissonanc e and the pil l: A n attrib utional approac h to<br />

study ing the arousal prope rtie s of dis sonanc e . Jo u rn a l o f Pe rs o n a lity a n d So c ia l Ps yc h o lo<br />

g y, 2 9 , 7 0 3 ± 7 0 3 .

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!