Relatively Fast! Efficiency Advantages of Comparative Thinking

21
tap'aidS/zfr-xgeiztr-xgelzfr0O109lztr210&d09z I xppws I S-=1 | 11l2Bl0Bl 11:55 | Art:2007-0096 Joomrl ol lj\pcrimcntill Ps)cholog\: Cenerrl 1009. Vo! . ... No .. l)0{) 000 Thonras Mussweiler Univc:rsity ol' Colognc 'I'he human nrind is a renrarkablc: cornparisonprocessor. When- cver infornration is perceived. proccsscd. ol evaluated, it scerns this inlirrrration is cornpared to a salient context, nornr, o[ stau- dard. Evcn the rnere percc:ptior) of'a physic:al objcct involves r comparison with a pertinent standard (Helson. 1964). The per- ceivccl size of a tirrgct circlc. for cxanrple. critically depencls on u'hetherthe target is surronndedby a set o1'largc or srnall circles. as is evidcnt in thc classic Ehbinghar-rs illusion (Coren & Enns. 1993). Similarly. the perceivcd rvcight of a target objcct depends on whether it is prc:sented with a set of heavy or light ob.iects (Brown. 1953). The pcrceptiou antl cvalualion of social targets secrns equally cornparativcin nature. Thc: perceivcd hostility of a target pl'l'son. lbr exarnple, critically dcpcnds on whether this pcrson is cvirluatcdin conrpurisonto a set ol'hostile or nonhostile persous (Herr, 1986). In much (hc same way, whellrer we sec a givc:nsocial issue as important (S. J. Sherman, Ahlm, Berman, & I-ynn. 1978).a trial judge as lenient (Higgins & Stangor. 1988), or ourselvesas competcnt (Morse & Cergen. 1970) all depends ou whether the pcrltinentcontext consists o1'a set of high or lclr.r' conrparison standards. Thc proclivity !o cngagrrin cornparisons is so rnuch parl und parcel o1' hunran infornratiou processing that L:\'en stinluli that arc not consciously perceived becausethey are 'I-hornrs ltlussrvcilcr. I)cpartment of General Psychology arrd Social Psl,chology. University of Cologne. Colo,rnc. Cennanv: Kai lipstudc, f)!'partlnentol l'sychulogy. Univcrsity of Illinois at Llrhana-Charnpaign- The present Leserrch rvas supported by a liuropcan Youne Irrvcstigator arvarrl lrorn the Lluropcln Science froundation. We rvorrld like to tlrank thc rrre mbcrs of the Würzburg Social Cognition Group and the ,Social Cogni- tion Cologne for stirnulating discussionsof this work. Correspondence corrcerning this article shoulcl hc atltlressccl to Thomas Musswcilcr. h)stitut fiir Allgemr:ine Psychologic und Sozialpsychologic, Univcrsität zu Köln. Groncwaltlstrasse 2. 50931 Köln. Gcnuany. Ii-rnail: thonras.rrrrssweiler@l uni-koeln.de :(trrq .\ilr(ricän P,\ch"l,'!icill'\s-rirctr,,il l+-15^)9/SI1.00 DOI: I0. I0-17/il{l(}1.1171 Kai Epstude University of lllinois at Urbana-Charnpaign presented subliminally are cornpared to a pcrtincnt standard (De- hac:ne et al., 1998). Ubiquityof Cornparison This tcndency toward comparative inlormation processing is striking because oi its renrarkable ubiquity. Psychological researcl.t has dcmonstrated horv dceply cornparisons pervade our thinking. For one, tlevcl<lpmental rcsearch shows that comparisons lrre a fun<lamenlalcognitivc skill that children ah'cady possessat a very rrarly age (Gentner & Medina, 199[ll Gcntner & Rattermann. l99l). Ilasic sirlilarities betrveen objects, lirr exanrplc. can be detcctedby chilclrenas young as 4.5 rnonths (Baillargeon, l99l ). Thc older chilclren get and thc rnore knowleclge thcy gain, the better they arc able to also detect less obvious sirnilarities. Fur- thcrnlu'e. resc'arch in dilferent areas of psychology shows that cornparisons play a key rolc in arcas as diverse as analogy (Cen- rncr & Markman, 1997). sinrilarity (Medin, Goldstone. & Gentner, 1993), categorization (Nosofsky, 1986; Nosollky & Palrncri, I997), stcreotyping(Biernat,2003), attitudes(Sherii & Flovland, l96l), person pcrception (Ilcrr. 1986; Higgins & Lurie. l9tl3; Smith & Zi.rate. 1992\, decision making (Choplin & Hurnrnel. 2002; Kahnernan & Miller, t986; S. J. Shennan,Houston.& Eddy. [999; Tverrsky& Kahncrnan. 1974), affect (Higgins, 1987). and the self' (Festingcr. 1954; Higgins. Strauman. & Klein, 1986; Miller & Prentice. 1996). Conrparisonsarc engaged evcn if thr:y arc not explicilly requested.In fact, it has bccu suggested that rvhcn stimuli of all levels ctl' complexity arc processed. they are spontancously comparecl to the l)onns and standards thg:y evoke (Kahneman & Miller. l9{16).When proccssinginfornrationabout anothcr person, ftrr exarnple, peoplc spontancously compare this person 1()thenrselves (Dunning & Hayes. 1996). Similarly. whcn processing inlirrrnation about themselves. people spontaneously comparc thernsclves to others (Festinger, l9-541Mussweiler & Rüter, 2003). Rcccnt cvidencc suggests that this tendencytcl makc o 0096 Relatively Fast! Efficiency Advantages of Comparative Thinking Corrrparisons are a uhiquitous process in inlirrmation processirrg. Seven studies eramine rvhether. how, rnd rvhen comparative thinking increascs thee{liciency ofjudgnrent antl choico. Studies 1-4 demonstrate thal procedurully prirningpalticipants to cngagc in rnore vs. lcss cornparistrn inllucnccs horv they process infonratiorr about a target. Specilically. theyretricve less infbrmation about the target (Studies lA. lB). think more about an inlbrmation-rich stanclard (Study 2) about which thcy activate judgment-relevant infonnatiorr (Study 3), and use this infonnation to compdusate tirr missingtarget infornration (Study4). Studies 2-.5dcm()nstrate the ensuing efficrency advartxlcs.Participants who are prirrred on comparative thinking arc faster in making a target.judgmcnt (Studies lA,2B, 4, 5) arrd have morcresiclu:rl processing capacitiers {irr n secondary task (Study -5).Studies 6 and 7 establish two boundary conditions by demonsftatiug that c(nnparative thinking holds efliciency advantages only if target antl standard are partl)'characterizcil by align:rble lcatures (Study 6) fhat aredifllcultto evaluate in isolatirur (Study 7). Thcse lindings indieatc tlrat corrrparltive thinking may oficn constitute a useful mcchanisrn to simplily infonnation prtrcessing. K cttr ortl s : compalison. cfficiency, thinking styles, procedural prirning ,\Q: 6

Transcript of Relatively Fast! Efficiency Advantages of Comparative Thinking

tap'aidS/zfr-xgeiztr-xgelzfr0O109lztr210&d09z I xppws I S-=1 | 11l2Bl0Bl 11:55 | Art:2007-0096

J o o m r l o l l j \ p c r i m c n t i l l P s ) c h o l o g \ : C e n e r r l1009. Vo! . . . . No . . l )0 { ) 000

Thonras MussweilerUnivc:rs i ty o l ' Colognc

' I 'he human nr ind is a renrarkablc: cornpar ison processor. When-

cver infornrat ion is perceived. proccsscd. o l evaluated, i t scerns

this in l i r r r rat ion is cornpared to a sal ient context , nornr, o[ stau-

dard. Evcn the rnere percc:pt ior) of 'a physic:a l objcct involves r

compar ison wi th a pert inent standard (Helson. 1964). The per-

ceivccl s ize of a t i r rgct c i rc lc. for cxanrple. cr i t ical ly depencls on

u'hether the target is surronnded by a set o1' largc or srnal l c i rc les.

as is evidcnt in thc c lassic Ehbinghar-rs i l lus ion (Coren & Enns.

1993). Simi lar ly . the perceivcd rvcight of a target objcct depends

on whether i t is prc:sented wi th a set of heavy or l ight ob. iects(Brown. 1953). The pcrcept iou ant l cvalual ion of socia l targets

secrns equal ly cornparat ivc in nature. Thc: perceivcd host i l i ty of a

target p l ' l 'son. lbr exarnple, cr i t ical ly dcpcnds on whether th is

pcrson is cv i r luatcd in conrpur ison to a set o l 'host i le or nonhost i le

persous (Herr , 1986). In much (hc same way, whel l rer we sec agivc:n socia l issue as important (S. J. Sherman, Ahlm, Berman, &

I-ynn. 1978). a t r ia l judge as lenient (Higgins & Stangor. 1988), or

ourselves as competcnt (Morse & Cergen. 1970) a l l depends ou

whether the pcr l t inent context consists o1'a set of h igh or lc l r . r '

conrpar ison standards. Thc procl iv i ty !o cngagrr in cornpar isons is

so rnuch par l und parcel o1' hunran infornrat iou processing that

L: \ 'en st in lu l i that arc not consciously perceived because they are

' I -hornrs l t lussrvc i l c r . I )cpar tment o f Genera l Psycho logy ar rd Soc ia l

Ps l ,cho logy . Un ivers i ty o f Co logne. Co lo , rnc . Cennanv: Ka i l ips tudc ,

f ) ! 'par t lnen t o l l ' � sychu logy . Un ivcrs i ty o f I l l i no is a t L l rhana-Charnpa ign-

The present Leserrch rvas supported by a l iuropcan Youne Irrvcstigator

arvarrl lrorn the Lluropcln Science froundation. We rvorrld l ike to tlrank thc

r r re mbcrs o f the Würzburg Soc ia l Cogn i t ion Group and the ,Soc ia l Cogn i -

t ion Co logne fo r s t i rnu la t ing d iscuss ions o f th is work .

Correspondence corrcerning this article shoulcl hc atlt lressccl to Thomas

Musswc i lc r . h )s t i tu t f i i r A l lgemr : ine Psycho log ic und Soz ia lpsycho log ic ,

Un ivcrs i tä t zu Kö ln . Groncwal t l s t rasse 2 . 50931 Kö ln . Gcnuany. I i - rna i l :

thonras . r r r rsswe i le r@l un i -koe ln .de

: ( t r r q . \ i l r ( r i c ä n P , \ c h " l , ' ! i c i l l ' \ s - r i r c t r , , i ll+ -15^)9 /SI1 .00 DOI : I0 . I0 -17 / i l { l ( }1 .1171

Kai EpstudeUnivers i ty of l l l inois at Urbana-Charnpaign

presented subl iminal ly are cornpared to a pcr t incnt standard (De-

hac:ne et a l . , 1998).

Ubiquity of Cornparison

This tcndency toward comparat ive in lormat ion processing is

str ik ing because oi i ts renrarkable ubiqui ty. Psychological researcl . t

has dcmonstrated horv dceply cornpar isons pervade our th inking.

For one, tlevcl<lpmental rcsearch shows that comparisons lrre a

fun<lamenlal cognitivc skill that children ah'cady possess at a very

rrar ly age (Gentner & Medina, 199[ l l Gcntner & Rattermann.

l99l) . I las ic s i r l i lar i t ies betrveen objects, l i r r exanrplc. can be

detccted by chi lc l ren as young as 4.5 rnonths (Bai l largeon, l99l ) .Thc older chilclren get and thc rnore knowleclge thcy gain, the

better they arc able to a lso detect less obvious s i rn i lar i t ies. Fur-

thcrnlu'e. resc'arch in dilferent areas of psychology shows that

cornparisons play a key rolc in arcas as diverse as analogy (Cen-

rncr & Markman, 1997). s inr i lar i ty (Medin, Goldstone. & Gentner,

1993), categor izat ion (Nosofsky, 1986; Nosol lky & Palrncr i ,

I997), stcreotyping (Biernat ,2003), at t i tudes (Sher i i & Flovland,

l96l) , person pcrcept ion ( I lcrr . 1986; Higgins & Lur ie. l9t l3;

Smith & Zi.rate. 1992\, decision making (Choplin & Hurnrnel.

2002; Kahnernan & Mi l ler , t986; S. J. Shennan, Houston. & Eddy.

[999; Tverrsky & Kahncrnan. 1974), af fect (Higgins, 1987). and

the sel f ' (Fest ingcr. 1954; Higgins. Strauman. & Klein, 1986;

Mi l ler & Prent ice. 1996). Conrpar isons arc engaged evcn i f thr :y

arc not expl ic i l ly requested. In fact , i t has bccu suggested that

rvhcn stimuli of all levels ctl' complexity arc processed. they are

spontancously comparecl to the l)onns and standards thg:y evoke(Kahneman & Mi l ler . l9{16). When proccssing infornrat ion about

anothcr person, f t r r exarnple, peoplc spontancously compare th is

person 1() thenrselves (Dunning & Hayes. 1996). Simi lar ly . whcn

processing inlirrrnation about themselves. people spontaneously

comparc thernsclves to others (Fest inger, l9-541 Musswei ler &

Rüter, 2003). Rcccnt cv idencc suggests that th is tendency tc l makc

o0096

Relatively Fast! Efficiency Advantages of Comparative Thinking

Corrrpar isons are a uhiqui tous process in in l i r rmat ion processir rg. Seven studies eramine rvhether. how,rnd rvhen comparat ive th inking increascs the e{ l ic iency of judgnrent ant l choico. Studies 1-4 demonstratethal procedurully prirning palticipants to cngagc in rnore vs. lcss cornparistrn inllucnccs horv they processinfonrat iorr about a target . Speci l ical ly . they retr icve less infbrmat ion about the target (Studies lA. lB) .think more about an inlbrmation-rich stanclard (Study 2) about which thcy activate judgment-relevantinfonnatiorr (Study 3), and use this infonnation to compdusate tirr missing target infornration (Study 4).Studies 2-.5 dcm()nstrate the ensuing efficrency advartxlcs. Participants who are prirrred on comparativethinking arc faster in making a target . judgmcnt (Studies lA,2B, 4, 5) arrd have morc resic lu:r l processingcapaci t iers { i r r n secondary task (Study -5) . Studies 6 and 7 establ ish two boundary condi t ions bydemonsftatiug that c(nnparative thinking holds efliciency advantages only if target antl standard arepart l ) 'character izc i l by a l ign:rb le lcatures (Study 6) fhat are di f l lcul t to evaluate in isolat i rur (Study 7) .Thcse l indings indieatc t l rat corrrpar l t ive th inking may of icn const i tute a useful mcchanisrn to s impl i lyinfonnat ion pr t rcessing.

K ct t r or t l s : compal ison. cf f ic iency, th inking sty les, procedural pr i rn ing

, \ Q : 6

tapraid5/zfr-xge/zfr-xgelzfr00109lz!r2108d09z I xppws I S-1 | l1lZBl0B | 11:55 | Arl: 2007-0096

]U USSWITII-llR ANI) IIPSTUDII

spontaneous conrpar isons rvhcn processing in lbrmal io l t about agivcn target is so strong that people even usLr conrpar ison standarcls

that-phcnonrcnological ly-are not cven pl 'Lrsent . bccause they

wcrc: prcscntc. l subl iminal ly (Mussu'e i ler & E,ngl ich. 2005; Mus-

srvei ler . Rüter . & Epstude. 2004a).

In a<id i t ion. the propcrnsi ty for cornparat ive processing is so

robust that compar isons arc evcn cngaged wi th standards that-

from a normatir,'c: perspective-arc unlikely to provide uselul

infornrat iorr about the targct . In one study demonstrat ing th is

robuslness (Gi lbert , Ci ics ler , & Morr is . 1995). part ic ipants com-pared their own per lbrnrance in a task to a sal ient other 's even

whcn th is person c lcar ly const i tuted an inappropr iate standard

because he had received addi t ional t ra in ing in the cr i t ical ahi l i ty . l

Further. abundant research has dcrnonstrated that on the level 01'

s inrple judgrncnts pcople evcn usc standards that a le c lear ly i r re l -

evant because they rvere selected at randorn. The nrost prornincnt

exarnplc for th is usc of random conrpar ison standards is g iven by' l 'vcrsky

and Kahneman's (1974) semirra l s tudy on . judgnrentalanchor ing-a phcnomenon that captufes l row cornpar ing a larget

to a nunrcr ic standard (c.g. . 657c) int luences subsecluent cst iuratcs

of the target . In th is c lassic study, part ic ipants cst i rnatcd the

pcrccntage o1'Af i ican nat ions in t l le Uni tcd Nat ions. Bcl i r rc doing

so. they considered a conrpar ison standarcl that was rant lornly

dctcnnincd by spinning a wheel o l t t r r tune. This randorn sclect ion

rnade i t c lcal ' to part ic ipants that the standards did not providc

useful in lbnnat ion about the cr i t ical target . St i l l , part ic ipants uscd

these stauclards as a basis l i r r their judgrnents. In lact , even judges

who are expcr ienccd expcl ts in the cr i t ical dornain arc in l ' luenced

by sr-rch randomly deterrn ined conipar ison standurds-even whcn

thc. juclges themsclvcs readi ly evaluate thesc standards as i r re levant(Engl ich, Musswei ler . & St lack. 2006). Why is th is (endency to

proccss in lbrrnat ion in a cornparat ivc mirnner so ubic lu i tous? Why

do pcoplc cornpare?

Eff ic iency of Cornpalat ive Thinking

We speculate that one reasrx 1br th is ubiqui ty is the efFic iency

of cornparat ive in l i r rnrat ion processing, People are ol icr t ovcr-

whehned by thr: abunclancc cll'cognitive tasks they have to nraster.

so thcy l iequent ly behavc as cogni t ive miscrs (Ta5i lor . l98l ) .

Consistent rv i th th is pr i lc ip le, atrundant research shows that peoplc

havc a strong tendency to rc ly on those infornrat ion ploccssing

slratcgies that are paf t icular ly c: f f ic icnt . For exarnplc, pec4r lc typ-

ical l l ' r t r ly on thc usc ofcategor ies in general (Bruner. 1957) and

stcreotypcs rn part icular ( IJodenhausen. 1990; Bodenhausen &

Lichtcnstc in. 1987: Macrae & Bodt- :nhausen. 2000; Macrae. Mi lne.

& Bodenhauscn. 199.1) becüuse doing so saves cogni t ive c lpaci-

t rcs. Sinr i lar l l , , pcople rc ly on heur ist ics that t ranslbrrn cornplex

tasks in lo s i rnplc . judgnrcnts because doing scl saves scalcc pl 'o-

ccssing rcsources ( ' [ ' r ,c lsky & Kahncrnan. 1974). I r t nruch thc sarnc

way, we rvould l ike to suggest that pcople as cogni t ivc rn isc ls nral '

rc ly on crompar isons because doing so s inrpl i f icrs inforntat ior t l l ro^cessing and fret:s scarc:c cognitive resources filr competing tasks.

Whv may th is br : the casc? Which psychological mcchanisnrs nray

conlr ibutc to thc proposed el ' l ' ic icncy udvantages of crrrnprr l t ivc

th i nk i ng?

To rnsrvcr th is quest ion. i t rs inrportant to takc a ckrscr lo<rk at

thc psychokrgical rnechanisrns that under l ic comparat ive th inking.

Any conrpar ison requircs thc psychological p l 'esenc:e of a compar-

ison standard. so lhat standard act ivat ion is a natural s tar t ing point

for (he analysis o1'conrpar ison rnechanisms (Biernat & Eidelrnan.

2007; Mussweilcr. 2003). Reccnt research dcnronstrertes that whcn

cvalr"rat ing a given target , judges spontancously act ivate standards

about rvhich they have abundant informat ion avai lable and that

thcy have frecluently used as standards in the past-so-callcrd

rout intr s tandards (Musswei ler & I l ,üter , 20()3; Rüter & Muss-

weiler. 2005). When askccl to evaluate their own cheerlulncss, for

exanrple. col lege studenl part ic ipauts spontaneously act ivated in-

fbrmation about the checrfulness of their bcst friend-probably the

staudard about which students havc thc rnost intonnat ion (Muss-

wt:ilr:r & Rütcr. 2003). Sirnilarly. when askcd to evaluate the

puilctuality of another person, palticipants spoutaneously activatr:ti

in tbnnat ion about the punctual i ty . r1 ' t l enrt" lu. . r -c lear ly the stan-

dard about which they havc the most in lornrat ion ( I )unning &

Hayes, 1996). I t is notablc that t l ' r is act ivat ion of in lbrntat ion about

the sel f is not l imi ted to spc:c i l ic in lornrat ion that pLr l ' ta ius di rect ly

to thc evaluatecl characteristic but also includes rnore general selt'

in l i rnnat ion. such as one's gendcr (Mussn,ei ler & Bodenhausen,

2002). Tlris suggests that a rather encompassing standard repre-

sentat ion is spontanc:ously act ivated (sec also Musswei ler & Rüter.

2003). This research dernonstratcs that cornparative thinking typ-

ical ly involves the spontaneous act ivat ion of in l i r r rnat ion-r ich stan-

dards for which part ic ipants havc anrplc . judgnrcnt-re lcvant in l i r r -

lnat ion avai lable.

The activation of such initlrnration-rich conrparison standarcls

suggests a f i rs t way in which conrparat ive t l t ink ing may hold an

el f ic iency aclvantage. l1 ' comparat ive th inking involves the act iva-

t ion o1 a standard lor which judges have anrple judgrncrt t - re levant

infornratior.r available. then they rnay use this standard informatiou

as a proxy for targct information that is unavailable or dif'l ' icult to

obtain. This not ion is in l ine wi th the suggcst ion by Medin et a l .(1993, p. 2-59) that "accessed in lbrnrat ion l i r r one concept wi l l tcnd

to bc carr ied over and tcsted l ( r r appl icabi l i ty to the other conccpt ."

In th is way. compar isons rnay al low . ludges to part ia l ly forgo the

sr:arch for sorne pieces of target inlrlrrnation alt<lgether. lnstead of

rctr icv ing al l the in lbrrnat ion about a target that is requircc l . . juc lgcs

may sirriply rely t>n rcaclily available inlirrrnation about the stan-

dard. When nraking judgrnents about thc: reactions of anothcr

person ( l toss. Green, & Housc. [977; for an overv icw. see

Kruegcr ' , 1998. 2007), l i r r example. onc nrav rc ly on rcadi ly

avai lable in lbrnrat ion abou{ onc's own react ions as a . juc lgntentalbasis. Using th is rcadi ly avai lablc: informat ion about thc sel f is

l ikc ly to require l 'ervcr cogni t ive capaci t ies than scarching for ,

rveighing. and integrating rclevant infornraticln about ths targc:t

pcrson. In (h is way, in l i r rnrat ion about a pcr t inent standard nray be

usecl to lill in lirr inlbrrnation that is nrore dil'l ' icult to obtain about

the judgnrcntal target (Medin c: t a l . , I993). I )o ing so is l ikc ly t<r

' Under spccific conditions, horvever, a comparison staudard may have

no influencc. [f a potential cornparisou standard is not perceivecl as being

rclated to a novel situation or evcn exnerienced ls a distraction frorn the

main task, t irr crarnple. it might have no eflect on an indiviclual's perfor-

rn r r rce in tha t nove l s i tua t i<n . Rescarch on l ra log ica l p rob lc rn so lv ing

found irrtriguing evidence for thc nonconsidcratitxl of potcntial contparison

s tandards whcn par t i c ip rn ts s i rnp ly d id no t rnakc a conncc t ion bc ts ,ecn a

descr ibed ; r tb lcm and a prob lem they have to so lvc (G ick & Ho lyoak ,

| 980) .

tapraid5/zfr-xge/zfr-xge/zfr00109lztr2101d09z l xppws l s-1 l 11l28l0} l11:55 l Art:2007-0096

CIOM P;\ RATIVII TFII N K ttr-G

sinrpl i fy the cr i t ical task in inrportar l t ways, part icular ly i1 ' re levant

infornrat ion about the target is d i l f icul r t r r obtain.

Once cornpar ison stat tdards are act ivated, t l ley arc cot lpared to

thc judgrncl l ta l target . PrOtninent conceptual izal ions etnphasize

that such couiparisotts are carried out alorlg rclaticlnal structures

between the individual attributcs of thL: targct and the standard

(Centncr & Marknran, ]rt)<)4, 199'7: Markman & Gcntner. 1993.

1996: Medin ct a l . . 1993). The basic assumpt iot . t of these structure-

nrapping approacl tes to ct tntpar ist ln is that in compar ing two

objects, pcople are inf ' luenced not tnerc ly by sets o l ' separatc

propcrt ies bu( a lso by how the propert ies arc i l l terrc latcd wi lh in the

standard and thc turgel. Jutlgc:s first detcrtnine a gcncral relational

structurc aniong indiv idual propcrt ics that is shared by the targel

and the standald and then pr i rnar i ly basc their cr lutpar ison on

propert ies that arcr re lated to thc shared or a l ignable re lat ional

structurc. That is . the compar ison focuses pr imal ' i ly on al ignable

leatures. whereas nonal ignablc l tatures are ignored. This pcrspc:c-

t ive on cotnpar isou is supported by a host of crnpir ical f indings

(fur an ovcrv iew, sec Gentner & Markrnan. 1997). For one. judgcs

typical ly consider lacts that are rc lated to the shal 'ed al ignable

stnlcture to bc rnore central 1() the comparisou than facts that are

not rc latecl to th is st ructurc (e.g. , Cl lemcnt & Gentner, l99l) ' Tht ts

cornrnonal i t ic ,s rc latcd to thc al ignable structule are tnore s l l ient

than thosc that arc unrc latcd. The salne appears to be t rue lbr

di f fercnces (Gcntner & Markrnan, 199'1: Markman & Gcntner.

1996). Whcn asked to l is t d i l l t rences betrveen a targct and a

standafd. peoplc tcnd to l is t a l ignablc: d i f ferenccs-di f t -erences

related to the common structure rathel ' t l lan t tonal ignable (unrc-

I atcd) c l i f ' ferenccs. Sint i lar ly . cotnpür i sons of consulnct prodtrcts

are of ten pr imar i ly bascd on al ignablc d i l lerences rvhcrcas t ron-

al ignablc d i f lerenccs are ignored (Zhang & Markmarl , 200l) .

With in the dotnain ot 'socia l iuc lgnrent ( l rere is abundant cv idence

denronslrat ing that s igni f icant othcrs are t tscd as standards to

cvalul te str rngcrs rv l to s l tarc at least some cornmtlnal i t ies ( i .e. , are

al ignahle) wi th the s igni i ' icant other (see Andersen & Chen' 2002'

for a revierv) . ' l 'hcse theclrct ical considerat ior ls and ernpir ical f ind-

ings hold that comparat ive lh inking involves a t i rcus on ä subset o l '

potcnt ia l l ) ' re lcr ,ant infornrx l i ( )n. Whcrt conlpal lng a targct to an

act i \ ,ated standard. judges sclect ivc ly lbcus on thosc targct l 'ca lures

that are rc latcd lo an al rgnable structurc and ef fcct ively ignorc

those targct lbatures that are unrelated 1() th is conlnlon structure.' Ih is

in lbrrnr t ional focus on al ignable lcaturcs suggests a sccond

wa-v in which comparative thinking rrlay hold el't' iciency advatt-

tages. Consider ing or t ly a subset of ' potent ia l ly judgnlent-re lcvant

targct informat ion is l ikely to require less processing capuci t ics

than consider iug a ntore etrcompassi t lg sct of iu l i r r tnat ion.

Takc l t oge thc r . t h i s r easo t r i ng i r np l i es t ha l co t r t pa ru t i v c

th inking involvcs i r t . f r .srnat ior t . f r tc t ts i t ' t thut i t a l lows i t tc lges t t r

l i rcus on a subset o1' potcnt ia l ly judgr l tc :nt- rc levant inforrnat ion

abou t t he . l udgn i cn ta l t a r l l e t . Compara t i ve t h i nk i ng a l so i n -

volvcs l r r /Trr t rn l io t l t , 'o t t , t ler . Conrpar isc ln is l ikely to i l lvo lve thc

s l ) on tandous ac t i va t i o l l o f an i n t t l r n ra t i on - r i ch con l [ ] i t r i son s tün -

dard. so that lvai lablc statrc lard inf 'or tnat ion nray bc t tsed as a

prox1, l i r r targct inform.r t ion that is unavai lable or d i l f icul t to

ob ta i n . I n comb ina t i o t t . t hesc t necha t t i sms t t f i n l b rn ta t i on l bcus

and infor tnat ion t ransfer are l ikely to cntai l ef f ic iency advan-

t agcs f o r co t t t pa ra t i v c t h i nk i ng .

Accuracy of Conrparat ive Thinking

Do these assumed ef f ic iency advantages colne at the cost o l '

accuracy'/ Does the locus on a sttbsel o1'relevant target information

and thc t ransler o l 'easi ly accessib le standarcl in lbrrnat i r ln lead tcr

less accurate . judgrnents? Not necessar i ly . In lact , i t seems that

bolh tnechani stls-iulirnnation I ocus and i nftlrnration transl'er-

will 01'ten lcad to judgmental outcolnes thal are contparable tt>

thosc of a less comparat ive ancl potent ia l ly more elaborate judg-

nlent p locess.

To scc this. let us first consider how inlbrrnation lbcus inllu-

cnces. juclgntcnt accuracy. ln l ine tv i th our assunrpt ion that infor-

mat ion {ocus dt>es not necessar i ly lead 1o lcss accurate j t tdgnlents '

previous rescarch has dentonstrated that time pressLlrc-a Iactor

inf- lucncing how ntuch judgmenrrelevant informat ion is consic l -

crccl-does not necessarily influcnce thc judgment per se. Specil'-

ically, putting p.uticipants under time pressure while tltey conlpare

an unknown qual t t i ty to a g iven nt tntcr ic standard inf lucnced only

1hc tirne participants neecled to generate a subsequent estimate. not

the ni r ture of th is est intate (Musswei ler & Strack ' 1999b). Fot '

exarnple, part ic ipants who were askct l whether the German r ivcr

Elbc is longer or shotler thun 890 knt subseqrtently madc: silnilar

estilnates o{ its actual length no matter whether they rvcrc given

only 5 s or ample time to tnake the conlprrison. 'l 'his

may bc thc:

casc bccause thc.iudgment per se is not printarily a function of thc:

amount of infornrat iot t tx t which i t is based, but of the i rnpl icat ions

of tl"ris inlil 'rnation. To the c:xtcrnt that the inlirrnlation generated

dur ing the f i rs t 5 s has s int i lar impl icat ions lor the length ol the

Elbc as thc intirrntatictn generatcd alierward. sitnilar estitnates

should result. In tnuch the salne way. the infornlittion foctts that

cornparativc thinking appears to c:lllail may not nccessarily lc:ad ttl

lcss accuratc jut lgntcnts.' Ihe

same secms to be the case lor infonnat ion t ransfer .

Whether using avai lable iufor t -nat ion about thc colnprr ison

staudard as a proxy lbr r . rnavai lable inforrnat ion about the

judgrnental target in l luences the accuracy o1' judglnent is pr i '

ma r i l y a l unc t i o t r o l ' t a rge t - s t anda rd s i n l i l a r i t y . 11 t a rgc t and

standard are highly s imi lar , then in l 'ormat ion t ransfer is t ln l ikely

to inrpedt- : accuracy, because thc t ranslcrrcd standard in lbrma-

t ion and the nt iss ing targct infbr lnat ion are l ikely to have

sinr i tar i rnpl icat ions. Ti r rget*stanclard s inr i lar i ty is one ot thc

rna in gu id i ng p r i nc i p l cs o1 ' s t anda rd ac t i va t i on (Fes t i nge r ' 1954 ;

Kahnen ran & M i l l c r , 1986 ) i n t ha t j udges t yp i ca l l y ac t i va te

standards that are s i rn i lar to the judgmental target . ' l 'o

the exlcnt

that th is pr incip le is in p lace, in lornat ion t ransfer rv i l l o l ien not

in l ' luence the outool t lc of a. judgrnent process per se.- fo

be sure. i t secnrs c lear that informat ion focus and intor-

rnat ion t ransier can both in pr incip le rcduce j r tdgment accul acy.

To the cxtcnt that important acct t racy-enhancing in l 'ornrat ion is

ignored or d istor t ing in lbrnrat ior t is int roduced. c(rrnpar i r t ive

thinking nray lcad to lcss accLrr i l te j t rdgnents (Chopl in & Hurn-

nrc l , 200-5). The above rcasoning. I rowever, suggests that th is

c loes not necessar i ly have l t l be, and may o{ len not be, the casc.

Wc invest igatcd hor.v contparat ive lh inking in l ' luences judgnlcnt

accLlracy in the c 'ontext of ' the present research. To do scl ' wc

cxamined whether thc judgnrcnt outconle depends on the extcnt

to which judges engagc in comparat ive th inking. For nrany

human juclgnrents-part icular ly those that are complex and rc ly

on iu l i r rnrat ion wi th ambiguous impl icat ions-object ivc l rucl

t ap ra id5 /z f r - xge /z f r - xge /z f r00 l09 l z f r2 l l 8 ,d19z l xppws lS :1111 /28 /08111 :55 lA r t : 2007 -0096

MUSSW[1II..l1R ANI) EPSTUDI]

va lues a re c l i l ' f i cu l t t o ob ta i n ( f o r a d i scuss ion , see e .g . .

K rug lansk i . 1989 ) . No n i a t t e r wh i ch c r i t c r i on l o r accu racy(K rug lansk i . 19891 Swann , 1984 ) i s used , howeve r . d i l Te rences

in judgnreut accuracy prcsuppose di l lerences in juc lgrnent out-

comes . I n t h i s r cspec l , e ra rn i n i ng whe the r . j udgmen t ou t co lnesdcpcnd on t he e x t r r n t o f compara t i v c t h i . nk i ng a l l ows us t c r

i nd i r cc t l y cxp lo re j udgn rcn t i l eeu l i . r c y i n s i t uu ( i ons i n r vh i ch t r uc

v a l u e s d o n o t c x i s t .

BoLrndary Concl i t ions

Our reasoning thus l 'ar i rnpl ies that cornparat ive th inking

invar iably lxr lds ef f ic icncy advantages. I t is inrportant to notc.

I towcvcr, thal the above analysis a lso suggests a f i rs t boundarycondi t ion lbr these cf f ic icncy advantagcs to occur. On the basis

o f a s l r uc tu ra l a l i gn rncn t pe rspec t r l c on compar i son (Cen tne r &

Markn ta r t . l . 994 , 199 ' l ' . Ma rk rnan & Gen tne r , 1993 , 1996 ; Mc -

d in e t a l . . 1993 ) . r ve have hypo thes i zed t ha t compar i sons a r c

ca r r i c c l ou t a l ong a sha red re l a t i ona l s tmc tu re a rnong i nd i v i dua l

l 'eaturcs of targct and standard. l f 'cornpar- ison requires such un

al ignablc structr"r rc. then conrparat ive th inkine lnay be hard to

cngagc and rnay consequcn t l y l ose i t s c f f i c i ency ad i ' an tages i n

s i t ua t i ons i n r vh i ch i r n a l i gnab le s t r uc tu re i s d i l f i cu l t o r i n rpos -

s i b l e t o es tab l i sh . ' l ' h i s

suggcs t s t ha t compara t i ve t h i nk i ng w i l l

p r i n ra r i l y ' sho rv c l l ' i c i ency advan tagcs when t he f ea tu r cs o1 '

targct and standarcl are al ignable ra lher than nonal ignable. In

the present research, wc examinet l how th is f i rs t potcnt ia l

boundary condi t ion inf lucnccs thc assunred el ' f ic icncy advan-

t ages o f compara t i ve t h i nk i ng .

A scc :onc l po ten t i a l bounda ry cond i t i on i s sugges ted by

Hscc ' s ( 1996 ) r csca rch on l hc d i l l ' e r c : r r t i a l eva luab i l i t y o f ' a t -

t r i bu t cs . Th i s au tho r has suggcs ted t ha t s t i r nu l us a t t r i bu tes

di f ter in thrr extrrnt to which thcy can bc crvaluatc:d in rc lat ivc

i so l a t i on . Fo r cxa rnp l c . c va lua t i r . r g r vhe the r a r non th l y r en l o f

€.50{) is a lo l or a l i t t lc for an apartmcnt in thc c i ty o1 Cologne

i s d i l f i cu l t l vhen t h i s a rnoun t i s cons idc red i n i so l a t i on . Thc

samc task becomes consic le lably casicr . howcvcr, when the

largct rcnt is conrpared to standald rc:nts o1's inr i lar apar lments.

Compara t i ve t h i nk i ng n ray bc pa r t i cu l a r l y e f l i c i en t i n s i t u l t i ons

in which a given at t r ibutc is d i f ' f icu l t to evaluate separately,

bccause i ts i rnpl icat ions beconrc apparcnt only whc:n considercd

w i t h i n t he who l c r angc r o f va l ues (Hsee , 1996 ) . ' I h c p resen t

resea rch a l so exam ined how eva luab i l i t y as a second bounda ry

cond i t i on i n f ' l u cncc rs t he c : t ' l i c i ency advan tagcs o f compamt i vc

t h i n k i n g .

' fhe Prcsent Rescarch

The preseut lesearch rvas designcd to exanr ine the existcncc.

nrcchanisrns. and boundarv condi t ions oi thcr ef l ' ic iency advan-

t ages o l co rnpa ra t i ve t h i nk i ng . l n t hese s tud ies , r ve app l i ed a

p rocedu ra l p r i n r i ng l og i c (S rn i t h , 1994 ) t o man ipu la te t hc e x t en t

lo u,h ich . judges re ly on cr l rnpar isons whcn processing target

inforrnat ion. Part ic ipants ' main task was to . judge an unknown

larget . Pr ior to t l lc cr i t ical judgnrcnt task, part ic ipants workt :d

< rn an os t cns ib l v un re l a tec l t ask t ha t w l s so l ved us i ng a mo re o r

l c ss co rnpa ra t i ve p rocess ing s t y l e . P t cv i ous l c sea rch suggcs t s

tha t such d i l l ' c r en t p roccss ing rno t l c s w i l l ca r r y ove r ( o t he

suhsdc lucn t l x rge I j udgn l cn t s {MusswL : i l e r . 200 1 ; t i r r an ove r -

v iew see Snr i (h. 1994). This procedural l ink bctrveen lhe pr i rn-

ing and the targct task al lows us to exanr inc the ef f ic icncy

advantages ol ' comparat ivc in l i r rmat ion processing by manipu-

l a t i ng t he ex ten t t o wh i ch j udgcs engage i n compar i son . I n l i gh t

ol the descr ibed ubiqui ty o l conrparat ive infurmat ion process-

ing, i t seenrs di f f icul t to creale s i tuat ions in lvhich judges do not

engage cornpar isons at a l l . As a conscquence. one can in l ' luenccr

only whcthcr cornpar isons are engaged to f l greater or lesscr

extcrut . A procer lural pr inr ing approach is an apt rvay ro nrani ; ; -

u late th is cxtr :nt of comparat ive in lbrnrat ion processing.

Studies l -4 examine l row comparat ive t l - r ink ing changes in-

lbrmat ion processing dur ing a . judgrnent task. Our rcasoning

suggesls that cornparat ive th inking changes how judges process

infon.nat ion in three nrajor ways. Speci f ical ly . wc conlend thatjudgcs rvho cngage in morc colnparat ive th inking (a) locus on

a subset of . and l i rn i t thc i r search for . judgrnent-re levaul targct

knorvledge, (b) act ivate inforrnat ion about a judgrnental s tan-

dard, and (c) use avai lable standard knorvledge as a proxy { i r r

t a rge t know ledge t l - r a t i s unava i l ab l c . S tud ies 1A and lB exam-

ine how p rocedu ra l l y p r i r n i ng compara t i ve t h i nk i ng changes t he

rvay i r r which judges scck infbrrnat ion about thc judgrnental

target . In part icular , we examine whether a l ' ter contprrr ison

pr iming part ic ipants part ly lbrgo the searcl . r fbr target knowl-

edgc: and thus search for and act ivate less targct knowledge than

do contro l part ic ipants. Studies 2A and -3 examinc: rvhetherjuclges rvho arc pr i rned on conrparut ive processing th ink rnorc

about a pert inent standard and act ivate more judgrnent-rc levant

infornrat ion about i t than do contro l part ic ipants. Study 4 cx-

amines whether act ivaled standard inforrnat ion is ascr ibecl lo

the iudgmcntal target .

Studies 2A.2B. r l , and -5 exanr inc the ei f ic iency advantages of

conrparat iYe th iuking nrore di rect ly wi th the help of two dist inct

rncasurcs. In a l l four studies, we cxaminc ' "vhether part ic ipanls rvho

cngage in rnore compalative processing are iäster in rnaking the

cr i t ical judgments. Funhermorc. in Study 5, we use a dual- task

paradigrn to exarnine rvhether participants wh<l engage nrore

strongly in conrpurat ive ploccssing require less cogni t ive resources

fbr the pr inrary judgnrent task and hi lve moLc: capaci t ics avai lable

fbr a secondary task.

The renrain ing two studies. Studies 6 and 7. exanr inc two

potrrntial boundary conditions ior the proposcd ell'iciency advan-

tages of cornparat ive th inking. Spcrc i l ical ly . Study 6 cxamir tcs

whether part ic ipants who are procedural ly pr i rncd on cotnpalat ivc

th iuking are faster in making target judgrnents only in s i tuat ions in

which target and stant lard share an al ignable structure that is casi ly

apparent . Study 7 lbcuses on cvaluabi l i ty as a second potent ia l

boundary condi t ion. Speci f ical ly . we c:xant ine wherher the el l ' i -

c iency advantages o{ 'conrparat ive th inking are apparent pr imar i ly

tbr cornpar isons of at t r ibutcs that are di f f icul t to evaluate in

i : o l a t i t - r n 1 l ou cva luab i l i t l t .

Srudy 1

Our f i r s t s t udy exam ined whe the r compara f i ve t h i nk i ng i n -

c lcecl changes in lbrmat ion processing in a lundanrental way. We

have suggestecl that comparal ive in l i l rmat ion processing may

a l l o r v j uc l ges t o l i n r i t t he ex t cn t t o wh i ch t hcy engage i n t he

t i rne- ancl capaci ty-consuming proccss o1 scarching lbr re levant

targct in l i ) rn lat ion. Becausc the use () f cornpar isons provides

tapraid5/ztr-xge/zfr-xge/zfr00109lzIr210'd09z I xppws I S=1 | 11l2Bl0Bl 11:55 | A(:2007-0096

CO MPA RATIVII TH I n" K In-G

juclgcs wi th intbrmat ion about a pert incnt standarcl , thcy c i rn usc

this lcadi ly avai lable in l i r rmal ior . r about thc stanclard as a proxy

lbr nr iss ing in lorrnat ion about the target . This impl ies that

comparat ive processing le ads to a more l int i ted scarc l t f i r r new

inforr lat ion whcn judging a novel ta lgct . Study 1 focuses on

one cen t ra l aspec t o f t h i s hypo thes i zed change i n i n l o rn ra l i on

proccssing by cornparat ive th inking. Sprrc i f ical ly , u 'c cxarninc

whetht : r judges rvho rc lSr nrorc hc:avi ly on comparat ivc th inking

do indecd search lor and act ivate less inforrnat ion about thcj udgn rcn ta l ( a rge t . To i nvcs t i ga t c t h i s poss ib i l i t y , we dcve lopec l

a paradignr that a l lowed Lrs to coulr t the picces of in lbrnrat ion

pa r t i c i pan t s r e t r i evc . Wc hypo thcs i zcc l t ha t co rnpa ra t i ve i n l i r r -

rnat ion processing would lcat l to a rcc luced scalch for infbnna-

t i ( )n about a targct .' f o

e xam inc t h i s poss ih i l i t y , w t r askcd pa r t i c i pan t s t o cva luu te

a target apart lncnI about which they rvere given a r ich database.

Part ic ipants ' t rsk rvas 1t l evaluate the apartment a long a nurnbcr

o f d i f f c ren t d imcns ions . Fo r each o f t hese d i r nens i i r ns . p i l r t i c i -

pants coulc l retr ieve background inforrnal ion f ronr lhe provided

database that a l lowecl them to niake a niorc infornted judgrnent.

I I ' con rpa ra t i ve p rocess ing i ndccd a l l o r vs j udgcs t o r nake t he

c r i t i ca l t a rgc t . i udgn ren t on t he bas i s o f l e ss i n f o r rna t i on abou t

thc targct . thcn part ic ipanls who werc proccclural ly pr imccl on

c:ornparat ive processing should retr ievt : less backgr<lund rn{br-

lnat ion about thc targot 1 ' ronr the database.

We e xarnincrd t l r is possib i l i ty in Studies lA and 1ts. In bot l t

s tudics. wc proccdural ly pr i r rcd part ic ipants to rc ly r r tore or ' less

heavi ly on cornparat ive processing beforc searching lbr infcrrma-

t ion and cr ,a luat ing the target apartn lcnt . Studies lA and lB usct l

d i f fcrcnt nrater i i l ls lbr the procedural pr iming of ' r t tore versus less

comparat ivc processing sty lcs. Ot l ler than that . proccdures were

cquivzr lc :nt .

Stud), 1A

Mrtluxl

Purt ic i lxrnts. Forty-scven studcnts rvcrc recrui ted r \ pxr t ic i -

pants. Thcy uerc cort tacted over the phone ancl asked t t t part ic ipatc

in a ser ies ol 'unrelated studics that would last for a tota l of 'aboul

I hr . As cornpcnsat ion. part ic ipants rvcrc o l lered €6 (about U.S.

$7 .50 a t t he t i r nc ) .

Protedure. Upon arr ival in thc lab. part ic iprnts werc led to

indiv idual cubic lcs and scated in f ront of a personal computer. Ncr

rnorc thau 3 indiv iduals part ic ipated in one session. The exper i -

rncnte r told participants that thcy would work on a ctlrtple of shot1.

unrc latccl s tudies explor ing human informat ion pr t iccssirg. They

first rvorkcd on a pap!'r-and-pencil procedural priming task. AII

prr t ic ipants were given sketchcs ol ' lwo sccncs th i l t werc takcl t

f rom Markrnan and Gcnl l )er ( 199(r) : onc depicted a wonran leaning

ovcr a table whi lc hold ing a cup ofcof fce. and the othcr depicted

a nran standing in t lont of a table rvhi le reaching for a borv l . F 'or

a f i rs t group of or"r r part ic ipants (n : 21.1. the two sketchcs wcre

prescntcd on thc san)c pagc. ' [ 'hese part ic ipants rvere expl ic i t ly

instructcd to corrpare thrr two pictu les and lo rvr i te down si tn i lar-

i t ics and di f lcrcnccs between thenl (comparat ive processing con-

di t ior) . For a second group o1'our part ic ipants (n : 26). the two

skelches rvere dcpic(ed on two separate pages. Tl tesc part ic ipants

rverc asked to dcscr ibc each of the indiv idual sccnes sc:Daratc lv(contro l condi t ion ) .

Upon cornplet ion ol ' thc pr i rn ing task. part ic ipants procecded

with the cr i t ical . iudgment task. which was adtnin istered on the

conrputer . Instruct ions cxpla ined that part ic ipants were about to

1nf tg part in a study of lnternet advert isentents. Wc explained

that a l though searching l t r r rn apartrncnt v ia the Internet has

becorne increasingly cornmon in our society. rc la l ively l i t t le is

known about which l 'acts are importanl to get an impression of

thc: apar{ment. We l 'ur ther in lbrmed par l ic ipants that their task

was t o j udge a f i c t i t i ous i r pa r tmcn t on seve ra l d imens ions . Thus .

they rnay, lor cxample: , be asked to iudge thc s ize of the

apart lnent . To bc in a posi t ion to make such a judgment. thcy

would reccive background in lormat ion 1 'or each ol ' the judg-

Inents thcy were askcd to nrake about the apartrnent . Part ic i -

panls wcre instructed to retr ieve as rnuch in l ' t r rmat i< ln as they

dcemct l necessary to lnake a . iudgment.Part ic ipants judged the larget apartn ient a long threc cr i t ical

d inrc:nsior . rs. Speci f ical ly . they est i lnated thc pr icc o l - t 'ent , thc

pr ice o1' ut i l i t ies. and the distancc to the unrvers i ty . Each o1'

thcrse judgments was presentct l on a scpal 'ate computer screen

along wi th a d i f ferent set of cucs represrrnt ing the avai lable

background informat ion l 'or thc g ivcn . jud-qment. A tota l ot ' 28

cues . 8 t o l 0 l b r each o f t hc . j udgn ren la l d i n rens ions , w r - r ' c g i vcn .

These cues were prescnted as labeled but tc lns on the cotnputer

scrcen. The but tous were vagucly labelc:d to g ive part ic ipants an

idea o1 what k ind ol in l 'ormal ion they representecl . For examplc,

tbr the judgnienl about the rent l i r r the apartn ient , the avai lable

cues includecl thc nurnber o1'bedroonts. the type of bui ld ing.

and the number o l 'bathroonts. Par l ic ipants could choosc to sec

the lu l l background infonnat ion that correspot. tded to each cue

by c l ick ing thc respcct ive bul ton on the scl 'een. To i l lustrate, for

the cr i t ical judgnrent about thc rent , part ic ipants were plovided

with l0 cucs. For one ol ' thc:rn the corresponding but t ( )n was

labeled roor i . r . L lpon c l ick ing on that but ton i t was revealed that

the apartnrent had four roorns. Part ic ipants could inspect th is

in lbrmat ion as long as thcy l iked and could then return to the

computer scrsen rv i th thc cr i t ical judgncnt and the l is t of

avai lable cues. Part ic ipants could l ' reely choose how many cues

they wantcd to select . und thus how nruch background in lbr-

mat ion about thc target apartment they rcccived. To provide

their ansrver, part ic ipants pressed an answcr but ton that was

plesented on thc sarne screcr l as the cr i t ical judgntent and the

avai lablc cues. They could then provide thci r atrswer on the

next screcn. Upon doing so. they procecded wi th the next

judgnrcnt . rvhich was prcsented on the subseqttcnt screen [o-

ge the r w i t h a ncw se t o f ' cues .

Result.s

We hypothcsized that part ic ipants who wcte pr imecl on a

lnorrr comparat ive processing sty le would search fbr lcss in lor-

mat ion about the judgnental targct than those who were pr i tned

on a lcss corrparat ive proccssing sty lc (contro l ) . ' Io cxarnine

th i s poss ib i l i t y , wc : su rn rned up t he number o f cues pa r t i c i pan t s

selected across the three . iudgments to rct r ieve background

infornrat ion about the target . Consistent wi fh our predict ion,

part ic ipants in a r l r l rc complrat ivc proecssing tnode retr icved

less backgrount l i r . r lurmat ion about the target judgnter t t (M ' -

tapraidS/zlr-xge/zfr-xge/2fr00109/zfr21OBd09z I xppws I S=1 | 11/28/08 I tt,Ql f l!.2902-oogq

MUSSWT]I I - IJR AND I ]PSTUDIJ

I n.l

1 3 . 0 - 5 , S D - 1 . 3 5 ) t h i r n d i d c o n t r o l p a r t i c i p a r r t s ( M - l 7 . l l ,

5 D = 1 . 0 9 ) . / ( 4 5 ) ' . - 2 . 3 1 , p : . 0 2 , d = 0 . 6 ( > . )

Srudy ll]

Merhod

Participttnts. Thirty-nine studenls of various tnajors at thc

LJniversitl, o1' Cologne pal'licipated in the study. 'l 'hey

rc:ccivcd a

chocolate bar as compensat lon.

Mueriuls unrl pntceclurc. Upon arrival in thc lab participants

were grceted by the cxpcr i rnel l ter a l ld scated in separalc cubic lcs.

Thcy were to lc l that they rvoulc l part ic ipatc in two separate (asks

that \ \ ,erc conrplete ly unrelatccl . The f i rs t task ( i .e. . the proceclural

pr i rn ing) was introduced as a pretcst to val idate mater ia l for a

study on pcrccpt ion. To el ic i t a comparat ive infoml l t t ion process-

ing sty le a f i rs t group ol ' thc part ic ipants ( r t - l9) was asked to

colnpare two parts of txre large picturc. 'Ihe

picturc showcd a coltlr

paint ing o1 a jungle scenc: rv i th a large nuntbcr of d i f fercnt ani tnals

(e.g. . a rnonkey, a sr lu i r rc l , a duck) and plants (c.g. . var ious t rees.

color fu l l - lowels) .3 hr t l re rn iddlc o l ' the picturc was a vert ical l ine

div id ing thc picturc into two equal ly s ized parts. Par l icrpants wcrc

to lc l to conrpare the tu 'o halves ant l wr i le dorvt l a l l s in l i lar i t ies and

di t t 'e rcnccs bctwcer) thern thcy coulc l f ind. In lhe contro l c t lnt l i t ion

(n - 201. part ic ipants reccivet l the satre p icr lure. but rv i thout 1he

l ine. 1-hey wcre askcd to dcscr ibe the piLr turc. This par l ( )1 ' the

cxper imcnt was a paper-ant l -pcnci l task.

Upon complet iou of the pr i r t t ing task, part ic ipants proceedcd

with the cr i t ical judgme nt task. in rv l t ich they wcrc again asked to

judge a (arge( apartn)ent . This task was s i tn i lar to thc one t tsed in

Study lA. Part ic ipart ts judgcd the apart l l lc :nt a long f ive di rncn-

s ions. nantc ly. t l tc overal l s izc of the apal ' tmL: l l t . the s ize of the

roonl that is for rcnt , the pr ice of t l te rent , the pr ice ofut i l i t ies, and

the c l is tancc to the univc: l 's i ty . A tota l o1 '4 '1 cues werc pl 'csontcd.r

For rrxample. the cucs for the price of the apartmenl included thc:

tota l nunrber of rooms ( i .c . . four) ancl the fact that the apartn lcnl is

locatcd in an oldcr hui ld ing ( i .e. . indicat ing an at t ract ive and

sonrcwhat largcr apartnrcnt) . arnong othel 's . The cues for the dis-

tancc to the univcrs i ty wcrc spcci l ic t ime est i tnatcs regardir lg how

krng i t would takc to leach t l tc ut t ivcrs i tv by using the t rant systcnl ,

r id ing a bikc. dr iv ing by car. or walk ing to the main canlpus area.

Judgrnents and cucs wcrc prcscntcd and arsrvers wcrc g iven in the

\ i l n ) c u i l ) i r \ i n S tu ( l y I A .

Result.t

We expccted part ic ipants who wcre pr inred r ln a morc colnpar-

at ive processing sty le to rc l r ieve less infornral ic in aboul t l le apart-

nrc:ut t l lan part ic ipants in lhe ct lnt ro l condi t ion. Tl terc l i r t 'e wc

calculatccl t l tc sutr i o l 'a l l retr ievcd cucs per part ic ipant , which

servccl as oLtr dependett t var iable. Thc: resul ts conf i rn lcd ot l r cx-

pcctation. Participallts who rvere prirncd on a l.nore colnparatlvejn l i r r rnat ion procrcssing sty le scatched lbr s igni f icant ly lcss in lbr-

nlirlion aboul the apartment (lll = 26.79. SD = 7.69) than did

pa r l i c i pan t s i n l hc con t ro l con t l i t i on (N l = 31 .30 . . tD :6 .31 ) '

1 (37 ) - 2 . . 01 . p - . 02 . d - 0 . 64 .

Distu.r .s iot t

Thc rcsul ts of 'Studies lA and lB suggest lhat cotnparat ive

proccssing indeed l i rn i ts thcr i tmount of target in lorr lat i tx thal

part ic ipants search l i r r bcfore nrakinc thc cr i t ical judgnrcnt Par-

ticipants who wcre prclcedurally prirtrecl to cngage in ä lll()re

comparat ive ntode of in lbrnrat ion processing v iewed fcwer p ieces

of in lorr lat ion about the target apartnrent than did controI part ic-

ipants.

Study 2

Our reasoning holds that judges who were prinretl to rcly morc

heavi ly on cornpxrat ive th inking are in a posi t ion to l imi t the

scarch for target knowledge as is apparcnt from Stucly l. because

cornparat ive th inking induces them to th ink Inorc about a l t

inltrrmation-rich cotnparison standard. Study 2 was designed to

cxant ine th is possib i l i ty . I f comparat ivc procc:ssing indeed induccs

judges to think ntore abortt a pertinent cotnparisolt standard' thcn

this srantlartl should be tnore acccssible subsccluetrt to the .iudg-lnrrut task. We used a lexical dccis ion task to see whether th is is

indeed the case. Participauts judged att unknowtt target City X for

which their hometown Wtirzburg constituted an information-rich

comparison standard. Il participants who were primed rvith com-

parative proccssing think nrore abor-rt this slanclard while iudging

City X, as we contend, then they should be faster lhan control

part ic ipants in recogniz ing Wti rzburg in a lex ical decis ion task.

Limi t ing the serrch f i r r target infor tnat ic ln in the way dc:mon-

sl rated in Study I is l ikely to render comparat ive i r t l t l r tnat ion

processing nrore ef f ic ient . Because searching 1or. evaluat ing ' and

weighing targct knor.vledge is an arduous task thal drarvs on

cogni t ive processing capaci t ies, the lcss targct in{brnrat ion part ic-

ipants scarch for , the laster they should be in judging the target .

Study 2 was also designcd to dcnronstrate that conrparat lve pro-

cessing rray al low judges to tnake largct iudgnlents n lore quickly.

Study 2A was designed wi th t rvo goals in rn ind: (a) to cx i tmi t tc

whsther lnore conlparat ive th inking indeed lcads to i l lcreascd

accc:ssibility of standard inlirrtnation and (b) to demonstrate for the

l i rs t t ime that more cor lpetrat ive th inking indced al lows judges to

rnake the cr i t ical . iudgments fastcr . Study 2B then l i )cuscs on Ihe

lat ter e l l ' ic iency advantage ()1 'compar istx by at tempt ing to repl i -

cale the expected f inding wi th a d i l1 'erent method to inc luce com-

parat ivc th i r rk ing.

Stuclv 2A

Participants were asked to litrnt an itnpression ol an unknorvn

City X and to subscclucnt ly tnake a ser ies of judgrncnts about i t .

City X was briefly described in a short paragraph in terms that

made i t gcneral ly conrparable to part ic ipants ' home ci ty o l 'Wt i rz-

burg. Whereas City X was generally cotnparable tcl Würz-but'g. we

: Un lcss o the i l v isc no ted , a l l repor led t tes ts invo lv ing hypothes lzed

differenccs arc one-tailetl. Although wc harl clear directional predictions

nrd rn rny au thor i t i cs recommend the use o f one- ta i l cd s ign i f i cunce tcs ts in

such cases . th is p l i rc t i ce i s cou tnrvers ia l (e .g . , Abc lson. 1995) . We adoPted

Abe lson 's (1995) recomt t tended compromise be tween the two cx t rc lne

pos i t ions in th is dcbatc . cons t ruc t ing a nu l l -hypothes is re jec t ion rcg ion o l

5 f l r in the theore t ica l l y expcc te t l ta i l and 0 . -5% in thc uncrpec tcd ta i l .

3 The jungle picture ruay be obtainetl frolrr the authtlrs.t Studies lA and ll l werc conductecl in difftrent cit ies. We adaptecl the

pafadigrn to fit spccitics of the local rcal estatc markcts. As I consequence,

thc number o f iudgnrcn ts and cues d i t f c red in ho th s tuc l ies

t ap ra id5 / f r - xge /z f r - xge /2 f r00109 /z f r2 lO8dogz l xppws lS :1 l1 l i 2B l08111 :55 lA r t : 2007 -0096

COMPA RATIV I i THI t -v K tn*G

also nrade i t c lear that i t was not ident ical to Würz-burg by stat ing

thal Ci ty X is located in a d i l ' ferent f 'ederal statc o1-Cennany. In

th is paradigrn, pal t ic ipants ' hornetown const i tutes an in lbrrnat ion-

r ich conrpur ison standarr l so that part ic ipants coulc l arr ive at theirjudgments nbor.r t Ci ty X b-v conrpar ing i t to Würzburg. To manip-

ru latc the extenl to rvhich part ic ipants rc l icd c ln cr tn ipar ison. wc

again used a procedural pr in i ing task. Wc askcd part ic ipants e i ther

to cornparc two picturcs in wr i t ing or to descr ibe thesc pictures

separately. I f coniparal ive processing holds c lTic icncy advantag-

cs-as we contend-then part ic ipants who were pr imed on rnore

c()mparat ive processing should be Iaster in. iudging Ci ty X.

Furthennorc. i l cumparat ive processing induces iudges to th ink

rnore about the in{ i i rnral ion-r ich conrpar ison s(audard. th is stan-

clard shoulcl becorne nx>re accessible. This, in turn, woultl be

apparent i f part ic ipants wcrc fastcr in recogniz i l lg thc standard in

a lcxrcal decis ion task (Ncely. 1917 ) . To exatninc th is possib i l i ty .

rvc asked part ic ipants to make a scr ies of lcx ical decis ions al icr

thcy had judgcd target Ci t ) , X. One of thc st imul i used in th is

lexical dccis ion task rvas the narnc o1' the pert iner l t L ' ( )mpar ison

stanr lard, the c i ty o l 'Würzburg. l f 'par t ic ipants who rvere pr imed on

more conlparat ive proccssing do indcccl t l r i r tk Inore about Würz-

burg u 'hen judging Ci ty X, thcn th is shoulc l bc apparcnt in shorter

re sponsc latcncies l i r r thc rvord Wi i r ; .burg.

M etltotl

Port ic i l tunts. ' l 'h i r ty-u j l le

studcnts at the Univc:rs i ly o{ Wi i rz-

burg rverc recrui ted as part ic ipants. They "vere approachcd in the

univc ' rs i t1, ' c l leter iu and askccl 'uv l te l l tcr they were wi l l i t tg to par-

t ic ipate in a psychological crpcr iment in exchatrge f i r r a c l rocolate

bar.

Pr<tceclttrc. Palticipants wcrc led to a separate room and seated

in separatc cubic lcs. ' l -he exper i rncnter cxpla ined that thcy would

wrl rk on sevcrral short exper i l r r rnts that were unrc la led to one

anothcr. -fhey

first worked on the paper-and-pcncil-based proce-

du ra l p r i r n i ng t ask uscd i n S tudy lA .

Af ter conrplct ion ol the pr i rn ing task, part ic ipants were each

scated in l ront 01'a laptop cornputcr to wt l rk on the os(cnsib ly

unrclated cr i t ical . judgl lgnt task. Instruct ions pointed out that par-

t ic ipants ' task was to lbnn an i rnprcssion ol ' an unknown Ci ty X

and to subsequent ly makc a ser ies of judgments about i t . Ci ty X

rvas then briefly clescribcrl in a short paragraph in tenns tha[ tnade

i t general ly conrparablc to part ic ipants ' honte c i ty o l Würzburg. In

part icular . Ci tv X rvas said to hc a univt : rs i ty town that rvas also a

ccuter o l ( icrnran Cathol icrsrn. In addi t ion i t rvas pointed out that

thc c i ty was färnous for i ts v ineyarcls ancl i ts scenic locat ion at thc

banks ol l large r ivcr . Al though th is inf i r r rnat ion rnadc Ci ty X

general ly cornparable to Würz-burg. wc also niade i t c lear that i t

wi ls not ident ical to Würzburg by stat ing that i t is located i r r a

di1l'ercnr {'ecleral state. Aftel participants had read the short para-

graph and had retlected on it for an additional 30 s, they prclcecded

to judgc thc: targel a long a tota l of f iv t : d inrensions. ' l 'hesc

dinren-

s ions pcrta inct l to the nunrber of inhabi lants, studct t ts , ntovie

thcaters. m.t jor br idgc:s. and strect car l ines. Part ic ipants were

instructed to makc these. ludgnrcnts as quickly ancl as accurately as

possib le. Euch of the targct qucsl ions \ \ ,as prcser l ted on a scparale

scrc:cn. Par l ic ipants were insl lucted 1() press the spacebar as s()on

irs they l la( i colne up rv i th an ansu'cr . Instruct ions cmphasizcd that

part ic ipants shoul t l not prcss thc spacebat ' bcf i r re they actual ly

knew their answer. Af ter pressing the spacebar to indicate that thcy

had corne up with an answer. a response lield appeared on the

screen and part ic ipanls were instructed to enter their ansrver intcr

this response lield. The critical response latencies wcre rccorded

from the lnoment the question appearccl on the screen until par-

ticipants pressed thc: spaccbar to indicate that they werrr Lcratly lo

glvcr all answer.

After cornpletion of' this .judgment task. participants proceeded

with the lexical dccis ion task, which rvas dcscr ibed as a tcst of

part ic ipants ' verbal abi l i t ies. Speci f ical ly , part ic ipants were lo ld

that thcy would be presented wi th a ser ies of let ter st r ings and

would be asked to indicatc whether each constitutcd a worcl of the

Gcrnran language. To indicate that the presented let tcr st r ing wes

a worcl. parlicipants were instrucled to press the riglrt crrnlrol

button (lnarkcd with a red dot). To indicate that the prcsented letter

string rvas a nonword. they rvere instructed to prcss the lcfl corrrrul

button (niarked with a grecn dot). Participants rvcre itrlilrtrcd that

i t was essent ia l lor th is task to makc decis ic lns as accurale ly and as

quickly as possib le. To al low thcnt to do so. they were fur ther

instructed to put thci r index f ingcrs on the respcct ive responsc keys

and to keep th is posi t ion thnrughout the task. Each let tcr st r ing rvas

prescntcd in the center of the screen unt i l part ic ipants had indi-

catcd tl"reir rcsponse. Eacl"r lexical dccision was followed by a break

of 1.000 ms bcfore the next letter stling was prcsented. This

scqucuce was rcpeatcd for each of the 15 let ter st r ings. Ten of

thesc: were German words (e.g., Munnel [marble] <>r Rulnnen

ltianel) ar]d 5 wcre nonrvords (e.9., polibal.f <tr legabip). The Iirst

l0 Iexical decis ion t r ia ls served as pract ice t r ia ls. rvhich were

included to hclp part ic ipants adapt to the task. I tems I l -15 rvere

the cr i t ical ones t l rat inc lucled thc target word. a nonword, and

three neutral rvords. The nanre o1'the pertinent standard \Viir:.lmrg.

rvhich servcd as our targLrt word. rvas prcsented at the I 3th position

in the scclucrnce. ' l 'he

neutra l words were prescntcd at the l l th,

l4th. and t5th posi t ions in the sequence. Al l i terns were presrr l l t r rd

in a fixed ottier that hud been randornly detc:rtnincrd.

Al ic : r cornplet ion of ' the lcx ical decis ion task. part icrpants rvele

thanked. thoroughly debriel'ed. and given tlieir cancly.

I le s u lts' l 'urget judg,rnenls. We cxpcctcd panicipants rvho were printed

ol t a lnore comparat ivc processing rnodcr to bc faster in n laking the

cr i t ical target judgrnc:nts than part ic ipants who were pr imed on a

lcss conrparat ivc proccssing mot le. To cxamine th is possib i l i ty , wc

analyzcd lhe rnean rcsponscr latencies to the cr i t ical . iudgntentsabout target Ci ty X (Clronbach's n . - .79). To contro l fbr oLr( l iers.

lbr cach response dimension rvc excluded response t imes that

dif'f 'ered by ulrre than 2 5'Ds lrorn the rneart (6.5%, of all responses:

Bar-eh & Charlrand, 2000) and averagecl tltcnt into one index.

lndccd. part ic ipants in a rnolc contparat ivc processing mode were

fastcr| in evaluating the target City X (.M : 8,494 ms, S/) = -59-5)

than contro l purt ic ipants (M = 12,299 rns, .SD - 1,248). 1(37) -

3 . 0 1 , 2 = . 0 1 . r i = i ) . 9 1 .

Wc also exalr inc:d whether part ic ipants in both condi t ionsjudgecl target Ci ty X di l ' f 'erent ly . Because t l te cr i t icul est imates

al low lbr a lnrost unrc\ t ra ined var ianse, we I l rs t excludcd al l cst i -

nrates tl)at clevirted liorn thc qucstion rnean b-v more than 2 S'I)s.

We thcn : - t ranslbrrned the indiv idual judgmcnts and averagcd

thenr into oncr index. Part . ic ipants ' judgtnc:nts of target Ci ty X did

tapraid5/zfr-xge/zfr-xge/zfr00109lzlr2108d09z I xppws I S:1 | 11l2Bl0B | 11:55 | Art:2007-0096

MUSSWITt l - l lR ANI) I IPSTUDII

not s igni f icant ly depcnd on u ' l re l l rcr thcy were pr inrcd on compar-

at ivc proccssing(M - -0.11, .9D : ( ) .121 or nol (M -- 0.14, J 'D - .

0 . 4 5 ) . / ( 3 7 ) . - 1 . 1 2 . n s .

Lexit'trl decisiot't.t. We hypothesizecl that participants who

werc pr inrcd on a nlorc conrpalut ivc pr()cc\ \ ino ntode would re ly

rnore heavily on infomration about thc pcrtiuent standard Wlirz-

burg when judging targct Ci ty X. I f th is is indeed the case, then

after the target judglnent . Wt i rzburg s l - rould be morc accessib le for

par l ic ipants whr l rvcre pr inrecl rv i th a more cornpurat ivc procr 'ss ing

rrrodc. As a consequence, rve expectecl fesponsc latcncies for

lexical dccisions abur.rt thc larget word Wi.ir:.burg to be shorter for

part ic ipants who r .verc p l inrcd to engagc in nr() rc comparal . iVc

proccssi ng.

We excluded response latencies t l rat d i l ' fcred by morc than 2

. l1fs l iorr r the mexn (5.8% o1'a l l rcsponses) f ronr our analysis(Bargh & Char l rand. 2000). Out l icrs were dist r ibuted across con-

di t ions. Anrong lhenr were the latcncies for three rcsponses to the

target rvord Wiiriburg. so the analysis oI thc lexical decisions is

based on the rcsponses ol 36 part ic ipants.

l ' l The rnearrs depicted in Figurc I arc consiste nt u. ' i th <lur reason-

ing. Those pilr'ticipants rvho had been inducccl to procLrss the targetjudgments in a nrore comparat ive rnanncr wele indeed i rster in

ident i fy ing the target rvorc l lVr l l r :Dro3 (M . . .688 nrs, SD - 161)

than contro l pal t ic ipanls (M - 8 l I nrs, 51) : 175). t (341 - 2 16r.

p - .02. r l - 0.74. h1 conlrast . no di f f 'crcnce crnerged in response

lalcncir :s to the ncutra l rvords (M = 638 Ins. S'D = 126 r 's . M :

638 n r s , SD - l l 6 ; I < l ) . I n a 2 ( co r rpa r i so t . t v s . con t ro l

pr inr ing) x 2 ( targcl word vs. ncutra l words) mixed-rnodcl anal-

ys is of 'varrancc (ANOVA), th is pat tc: rn was borne ou( in a s igni l ' -

i can t i n t e rac t i on c l l ec t , f ( 1 . 34 ) . - 8 . 8 t t , p , , , , , . ( ) 1 . pu l t i a l r l 2 : . l l .

ln th is analysis thc nrain c l lect o1'word type also reached s igni l ' -

icancc, 1. . (1.34) ' ' - 29.60, p " ' .001. part ia l r t r - . -17. indic i r t ing

that part ic ipants r .vere fast t : r in rcsponding to the neutra l words

than to thc targct rvord. The rnain e lJ 'ect of pr i rn ing fa i led lo reach

s ign i f i cancc , F (1 . 3 i 1 ) : 1 . 91 , n . s .

Srudy 2l]

Wc dcsignccl Stucl l , 2B to repl icate the eiTic iency advantages of

cornparat ivc processiug dcnrol ls t rated f i r r the f i rs t t inrc in Study

2A. using a di l lcrcnt task to induce di l ' f 'crcnt ia l levcls o l compar-

at ive processing. We aguin used a procedural pr iming logic to

induc:c a n lorc versus less conrpatat ive processinr sty le, th is t ime

with a d i l lcrent and nrore indirect pr i rn ing lask. Pr ior to the cr i t icaljudgment task. participants rvorked on an unrelated prirrring task in

which they had to l ind c ight errors in a drawing of a c i rcus scene.

For part ic ipants in the cor lparurt ivc proccssing condi t ion. th is

priming was a conrpafative error scarch that took thc fbrnr of an

or ig inal and copy task. Speci f ical ly , these part ic ipants werc pro-

v ided wi th two cartoons of the sarne scene-ulr ' l r r ig inal" and a"copy"-thxt di11'ercd u,ith respect to t:ight dctails. The cartoon

depicted a c i rcus scenc wi th c, lowns. t ightrope ar t is ts. spectatol 's .

and aninral tanrels. Part ic ipants ' task u 'as to ident i fy thc cr ight

c leta i ls wi th rcspcct to rvhere the or ig inal and thc copy di f ' l 'ered. To

solvc this task, participanls thus had to repeatedly comparc thcr two

pictures. For part ic ipants in thc contro l condi t ion, howevcr. the

prining rvas an absolute error search. 'l 'hesc

par(icipants werc

provided only with the original caltoon of the circus scene and

rvere instructed to identify eight dctails that constituted logical

inconsistencies (e.g. . ra indrops underneath an unrbrel la) . To solve

this task, part ic ipants d id not have to engagc in c()r rparr t i ! 'c

processing. I four reasoning is correct , then part ic ipants who solve

the pr iming task by repeatedly conrpar ing the two pictures shoulc l

rc:ly mtxe hcavily on comparisons in the targct task. rvhich should

in turn al low thenr to colnplete the target lask morc quickly.

Method

Part ic i l tut t ts . Forty stuclents part ic ipated in the cxpcr i lnent .

They wcrc: approachcrd in thc univcrsity cafeteria and askc:d lo

part ic ipatc: in a br ic f ser ies r . r1 psychological studies. Part ic ipants

wcrre oll'ered a chocolate bar as conrpcusation.

Pror:etlura. Upon agreement to (ake purt in the study. partic-

ipants rvcre led to a separate room. greeted by the expcrimenter.

ancl seatcd in separate cubic les. The exper i rnenter expla ined that

lhe participants' task was to work on seve ral short experitnc:ttts tlrat

wcrc unrelated to one anothcr and rvere adrninistercd togethcr

solely lbr elTicicncy reasons. They first worked t>n a paper-ant1-

penci l procedural pr i rn ing task. A f i rs t group oI our par l ic ipant \

1n : 201 rvas in the cornparat ive processing condi t ion. They rvere

givcn two very detai lcd sketches ds:pict ing the sar lc c i tcLts scenc.

Onc picture was labeled as the or ig inal the other as the copy.

Part ic ipants were instructed to c loscly examine the two scencs and

to idcnt i fy e ight detai ls ("c: r rors") thut d ist inguished the copy t ront

the or ig inal . Thcse dctai ls had to bc c i rc led in the cnpy. Thc secoud

group ot our part ic ipants (n : 20) was in the contro l condi t ion ancl

rc:c:c:irrc:tl jusl onc: picturc (narnely, the original liorn the cornpara-

t ive processing condi t ion). Thesc part ic ipants were instructed to

t ' ind eight logical rn istakes (e.g. , ra indrops undemeath an r . rmblel la)

and c i rc le thenr. These logical mistakes wcre the same detai ls that

dist inguished the copy f rom thc or ig inal i t . r the comparat ive pfo-

ccssing condi t ion.

Af ier conrplet ion of ' the pr inr ing task. part ic ipants were each

sLrated in front of a lapt<4r conlputer to work on the "uurelated"

cr i t ical judgnrent task. which was ident ical to thc onc usecl in Study

2A. ' I 'hus. part ic ipants again fornrcd an impression ol target Ci ty X

and subsequent ly rnade a scr ies o1 judgments about i t . A{ ter

cornplc: t ion of the judgnrent task. part ic ipants were thankcd. thor-

oughly debr icfcd. and of fercd u bar of 'chocolate.

1 000{,

o - ^ ^6,) / UU' :

600

G

u - ^ ^E J U U

o z u uoo 1 0 0d

0

lr i".n"i i i i ;- l

l"ä:::;i:"' ]

neu t ra I standa rd

Figurt ' l . In Study 2A, mean response latcncies for lex ical decis ionsrhout thc pcrtinent starrdard versus ncutral u'ortls by priming (comparativc

AQ:7 \ 's . contro l ) . [ ] r r0r bars reprcscnt stant lar t l crrors of t l re rnean.

tapra id5/zf r -xge/zf r -xge/z l r00l09lz I r210Bd09z Ixppws IS 1 | 111281081 11:55 |Ar t :2007-0096

COMP;\RATIVI: | THINKING

Re.sul ls

Wr' cxcludetl response tirnes that dif f'ered hy nrore tlran 2 5/)sl ronr thc nrcan (( r . I 7c o1'a l l responscs) l ronr fur ther analysis (Bargh

& Chartrand, 2000). Furthcrnorr : , we averaged par l ic ipants ' rc-sponse latcncics to thc f ive cr i t ical targct c lucst ious into onc index

cl l processing spced.

l f cornparat ivc in l i rnnat ion processing does indecd hold the

assunrcd ef f ic icncy advantages. thcn part ic ipauts in thc cotnpara-

t ive proccssing condi t ion should be faster in uraking the cr i t ical

targct . judgrncnts than part ic ipants in the absolutc procL'ssing c(rn-

t l i t ion. We analyzecl part ic ipants ' tneau responsc latcncies fbr t l re

cvaluat ion of the ta lgct Ci ty X lCronbach's r t - .69) to cxanr inc

this possib i l i ty . Indecd, part ic ipants rvho processccl the cr i t ical

. iudgnrents in a rnore cornparat ive rnanner rverc faster in . judgingtargct Ci ty X (M - 6.974 nis, SD : 503) than contro l part ic ipants(M = 9.431 rns.5 '1) = 8f t4) . / (38) : 2.42. 1t : . ( )1, d = 0.71 .

Aca in , n ' e exan r i ncd r vhe thc r pa r t i c i pan t s i n bo th cond i t i ons

. judged targel Ci ly X di f t 'erent ly . ' l -o t lo so. rvc excluded ust i -

mates that deviatct l f ronr the quest ion nrcan by rno|e than 2 Sl)s.

i - t r i - rnsfbrnled the indiv idual . judgnrcnts, and averaged thern into

onc i ndcx . Pa r t i c i pan t s ' . j udg rne r r t s t l l ' t a rge l C i t y X d i d no t

dcpcnt l on whether thcy werc pr inrecl on cornparat ive process-

i ng (M - - 0 . 1 - s , SD : 0 . 40 ) o r no t (M : 0 . 1 - s . SD : 0 . 62 ) .

/ ( 38 ) = 1 .8 ( ) . n . s .

[)i.scu.s.tio tt

Thc resul ts o l 'Studies 2A and 28 havc at lcast threc iurD()r tant

imp) icat ions. Fi rst . the f indings of Stut ly 2r \ f i r r t l rcr spcci ly how

conrpar ison changes infr . r rnrut ion proccssing in a u 'ay that rnay

al lou ' fbr e l ' l ' ic ienc:y advantages. Stu( ly I dcmonstratcd that com-parr t ivc informat ion processing l inr i ts the amount of informat ion

. judgcs act ivatc about the. judgmental targct . Study 24 supplcnrcnts

th is l inc l ing by hint ing at thc process that nray have alkrwcd

partrcipants in a coniparative proct:ssing rnodr: to partly ibrgo

act ivat ing (arset knowledgc. Speci l ical ly , the fact that exper in len-

ta l part ic ipants recognized Würzburg nrore c lu ick ly in a lcx ical

dccis jon task sr . rg_ecsts that thcy had more extcnsively thought

about and used in l 'ormat ion about th is pert inent standard than had

con(rol part ic ipants. C- 'onlpar ison may thus al low judges to l i rn i t(hc scarch l t r r judgrncnt-re levant targct knowlct lgc by inducing(hern to focus morc on re:adi ly avai lable standard knorvlct ige.

Sccond. thc f indings o1'Studies 2A and 28 der lonstrate that

par l ic ipants who lvere pl imed on lnore colnpaf i l t ivc pr t rccsr ing

rvere able to rnakc thc clitical judgrnents faster than control par-

t ic ipants. The f inc l ings of Study 2,4 provide in i t ia l suppor l lbr the

not i ( )n thal comparat i \ re informat ion proccssing indcccl has c: f f i -

c iency advantagcs. in that t l . rc usc ofconrpar isons nray speecl up the

. judgrnent proccss. Thc f indings of Study 28 provide lur thcr sup-port lor th is idea. Again. part ic ipants rvho werc procedural ly

prinred t<l rely morc hc:avily on cornparisor.rs when processing the

cr i t ical target judgrncnts wcre ablc 1o makc thesc judgrnents faster

than contro l part ic ipants. ' fh is

repl icat ion o[ ' the cf f ic ic :ncy advan-

tage in i t ia l ly dernonslrated i l Stut ly 2A was obtainc:d r . rs ing a

di l ' tcrent pr i rn i r rg nre(hod that int lucccl thc a l lcr la l ivc pr()cc\ \ ing

stv lcs in a rnolc indrrcct rvay ancl thus cmphasizes thcr gcncral iz-

ab i l i t y o f t hc ob ta i ned l ac i l i t a t i on c l l c c t .

Thir r l . thc presr 'nt resul ts do not providc any indicat ion that

thcsc cf l ic icncy advantagcs o1-conrparat ive th inking conre at the

cost of judgnrent accuracy. In l ight o1 the fhct that target judgnients

pcr la ined to a I ' ic t i t ious c i ty , o l course, the accuracy of thescjudgrnc:nts cannot be assessed. Di l ' f 'erences in accuracy. however.

prcsupposc lhat target .judgnrcnts by the cxperimental and thc

control groups differ. T'his does not appear to be the casc.

Stucly 3

Study 2A provides in i t ia l ev idence suggcst ing that conrparat ive

processing induccs judges to think nrore about an inlorrnalion-rich

conrparison standard. This is apparcnt in the tact that judgc:s

recognized thc lume of the standard morc cluickly in a lexical

dccis ion task. Our reasoning, however, holds not only that judgcs

think about a pert inent stanclard but a lso thal thcy rrct iv l tejudgment-re levanl infornrat ion about i t . This act ivatcd standard

inlormation can then be used as a proxy lbr target infornration that

is not reacl i ly avai lable. In th is respect , the spontaneous act ivat ion

of standarcl information rnay contribute to the elf iciency advan-

tages of cornparat ivc th inking. We designcd Study 3 to providc

additional supporl for the spontancous activation of standard in-

Itlrrnation in the realnr o1 pcrson judgmcnt. Prcvious research has

denronstrated that when judging others, pcople often use them-

selvcs as a cornparison standard and spontancously activatc inlilr-

malion about their own standing 6n llgr judgrnental dirncnsion

(e.g. , Dunning & Hayes. 1996; Musswei ler & Bodenhauscn,

2002). When judging how punctual another person is . f r r r examplc,

peoplc spontancously activate inlbrnration about their own pr.lnc-

tual i ty (Dunning & Hayes. 1996). The prcsent analysis holds that

the nrorc purt ic ipan(s engagc in comparat ive th inking. the nrore

in{ 'ornrat ion about the judgnrental s tandard shoulc l bc act iv i r ted.

S tudy - l was des igned t o exa r r i ne t h i s poss ib i l i t y . To do so ,

wc asked our part ic ipanls to fc l r rn an i rnpression of and nrake a

iudgment about a br ic f ly descr ibed target person. Spcci f ical ly ,

part ic ipants f i rs t l i r rnred an impression aboul a person who was

descr ibed as engaging in a uumber o l ' tasks that are re latcd to

manua l sk i l l s ( c . g . , r cpa i r i ng a b i ke ) . Pa r t i c i pan t s t hen . j udgedhow manua l l y sk i l l ed t h i s t a rge t i s . P r cv i ous resca rch (Dunn ing

& l {ayes. 1996) suggests that the scl l is used as a stant lard in

rnaking thcse judgrrc:nts. Subsequent ly, part ic ipants rverc : rskcd

to ei ther descr ibc an aspect that is < l i rect ly r r : lated to thc sel l as

the pe r t i ncn t s t anda rd o f t hc p reced ing . l udgmen t ( e .g . . a r nanua l

sk i l l t ha t t hey {hc rnse l ves posscss ) o r un re l a tec l t o t h i s s t anda rd(e .g . . an advan tnge o l ' hav i ng manua l sk i l l s ) . I f ou r r eason ing i s

co f rec t , t hcn t he n ru r c pa r t i c i pan t s ensagc i n co rnpa la t i ve t h i nk -

ing, thc nlore t l lcy s l rould spontaneously act ivatc infornrat ion

abou t t hc i r own manua l sk i l l s wh i l c j udg ing t he t a rge t pe rson .

As a conscquence, part ic ipants who had been procedural ly

pr i rned to engage in conrparat ive th inking should be laster in

d e s c r i b i n g o u e o l ' t h e i r o w n m a n u a l s k i l l s . N o t a b l y , t h i s

speed-up shor.r ld be apparcnt only lor descr ipt ions that re ly on

inforrrat ion that is re latc:d Io the colnpar isnn standart l ( i .e. . thc

scl l ' ) . Standard-unrelated i n l 'urmat ion ( e. g. , advantagcrs r r1 ' hav-

i ng manua l sk i l l s ) shou ld bc cc l ua l l y acccss ib l e ( i r r pa r t i c i pau t s

rvho rvere pr inred to eugage n)orc hc:avi ly in cornparat ive th ink-

i ng and con t ro l pa r t i c i pan t s . so t he desc r i p t i ons l ha t r c l y on

standard-unrc lated in lornral ion should bc given equal ly I 'ast by

both groups.

I t )

tapraid5lztr-xgelzfr-xge/zfr00109lztr2108d09z I xppws I S:1 | 11128108 | 11:55 | Art: 2007-0096

M USSWITII-LIR ANI) IIPSTUI)II

Method

Part ic i l tnt t t .s . We recrui tcc l 65 undergraduatcs at thc Univer-

s i ty o1' Colognc- as part ic ipants. They were approachcd in a uni-

vers i ty cal 'eter ia. asked to part ic ipate in a ser ies of ' unrc lated

psychological studies, and ol lered a chocolate bar as compensa-

uon .

Muteriul.r ttrttl proteilure. On arrival in the lab. participants

rverc grcctccl by thc expcr inrcnter and lcd to a cubic le that was

equippecl wi th a personal conrputer . Part ic ipants lvcre informcd

that thcy woul t l r .vork on a scr ies of unrelated s lucl ics that were

adnl in istercd t ( )gethel solc ly to savc scal 'ce rcscarch rcsources.

Thc: rnater ia ls ant l procedures used in Study 3 wcrc s imi lar to

thosc crnploycd by Musswei ler ancl Bodcuhauscn (2002). The

expcr imentcr f i rs t cxpla incd that bccause the second stut ly in-

volvcd a sornelvhai conrplcx exper imental procedurc, he would

cxplain i t to thc part ic ipants beforehand. The expcr imenter then

proceedecl to expla in that the secclnd study consis lecl of a scr ies of

t r ia ls wi th th lee di l ' l t rent tasks cach. For each t r ia l , part ic ipants

would f i rs t be cxposct l to the descr ipt ion ol 'a target person and

would be asked lo lonn an inrpression of th is person ( i rnprcssion

fornrat ion task). Once thcy had l i r r rncd an imprcssion. they would

then be instructcr l tc l judge th is person on a dinrension that is

rc lated to thc informat iorr providcrd ( judgment task). ' l 'o

do so, thcy

shotr ld f i rs t forrn their . judgment in their nr ind and press the

spaccbar once they had donc so. Thc judgrncnt should thcn be

entcred on thc ncxt con)putcr screen (1or a s i r -n i lar procedure, see

J. W. Shcrrnan. Klc in. I -askey, & Wyer, 1998). Subsequent to th is

. judgnrcnt . they would be asked to c lcscr ibe a concrcte ent i ty that is

re lated to thc tafget descr ipt ion (dcscr ipt ion task). Again, they

should f i rs t lorr l th is t lescr ipt ion in their rn ind and press the

spaccbar oucc thcy had c lone so to then cnte[ th is descr ipt ion on t l re

next corrputer scrcen. For c: rample. thc:y nr ight f i rs t lcccivc a

dcscr ipt ion deal ing u, i th horv a target pcrson has at l justed to l i fe in

Colognc, then bc asked to judge horv wcl l th is persou lus adjusted

on a sca f c l l ' on r I ( rm t a t a l l ) 69 ( ve r j ,we l l ) , and f i na l l y be askec l(o c lescr ibc a concrctc advar l t i lge oi l iv ing in Cokrgne. Again, they

rvcrc instructed to t l rs t th ink ol ' thc cr i t ical ent i ty (e.9. , advantage

of l iv ing in Cologne) arrd to press the spacebar as soon as soulc-

thing had conlc to their rnincl. They rvere fu(her inlbrmed that ive

rvould asscss responsc laterrcics lor the third lask in cach trial. so that

thcy should tr1 to providc their answcr as fast as possiblc. For the first

t \Ä ' ( ) tasks. no rcsponse latcncies would be assessecl so that they

could take their t inre tu rvork on them. To bc able to provide thci r

ans\1 'crs as last as possib le lcrr the th i rd task, they should put their

r ight indcx f inger on the spacebar and kecp th is posi t iou through-

out the cxpcr iment. F inal ly . part ic ip i lnts rvcrc g ivcn the opportu-

ni ty to c lar i f i r aspccls o l the expcr imental proccdurcs. The rctnain-

dcr of the L:xprrrimenl rvas se lf-irdnrinistered.

Par l ic ipants were I ' i rs t handed a fo lder that inc ludcd a procedural

pr i rn ing task s imi lar to thc onc uscd in Study lB. Al l pat ' t ic ipat t ts

received thc sanrc junglc p icturc used in Study lB. This l inrc. wc

acldccl f i lur f ine l ines ol 'ahout I crn euch to the picture to nrark the

ccnter of each ol ' thc four s ides. This a l lowed part ic ipants in the

c:xpcr imcntal gr()up to ident i fy the di f ferent halves oi the picture.

Expcr imental part ic ipants werc askccl to comparc thc: lc : f t and r ighl

halves of thc jungle scene. CoDtrol part ic ipants were askcd lO

sirnply dcscr ibe thc sccne.

At'tcr corrplction of this f irst part ol' thc sl,udy, participauts

procccded rvith thc seconcl pirt, rvhich rvas adrninistercd hy the

computer. Participants workecl on a total of three trials that fol-

lorved the descr ibcd sequence of impression lbrmat ion. targctjudgnrent , ancl c lescr ipt ion task. For each t r ia l , the l i rs t screen

instructed participants to caret'ully read the subsequent tarllet

dcscription and to lbrrn an implcssior.r of thc describc:d person. Thcr

ncxt sc l 'cen renr inded thcrr that responsc latcncics would not bc

assessed l i r r th is i rnprcssion l i r r rnat ion task so that they could takc

their t ime. Part ic ipants werc then exposcd to a descr ipt ion of

a target pcrson and rvere asked to lbnn an inrpression o1' this

person. They were instructed Io press the spacebar oncc they hud

done scl. The ncxt screen again renrinded them that response

latencies would not be asscssed for the subsequcrnt judgment task.

Part ic ipants wcre thcu to. judge the target person on a 9-point scale.

Again, they rvere told to press tlre spaccbar as soolr as they had

formed their .judgment. After reporting this judgrnent on thc next

screen. they wcre renrinded that for thc descriptioil task rve rvould

assess their response latcncies so they should respond as l is t as

possib le. Part ic ipants were then instructed to th ink o1'an ent i ty that

is re lated to the target descr ipt ion. to press the spacebar as soon as

this ent i ty had comc to their mind. and 1o then dcscr ibe i t on the

subsequent screen. This cxact sequel lcc was repeated lbr a l l t l r ree

tr ia l s .' fhe

f i rs t two t r ia ls were inchrdcd to obtain a measure of par-

t ic ipants ' basel ine speed of responding. For thesc t r ia is the dc:-

scl ipt ion task was unrelated to thc scl t 'and did thus not assess thL:

accessib i l i ty o1'knowledge about a potent ia l s tandard subsequcnt

to person judgment. Morcr speci l ical ly , in the f i rs t t r ia l , part ic ipants

formed an inrpression aboul a person who had moved to Cologne

2 years ago and hacl nr ixed l 'eel ings about l iv ing in a b ig c i ty . They

thcn. fudgcd how much the target person l ikes l iv ing in Cologne

and f inal ly descr ibed a concrete disadvantage ot ' l iv ing in Cologne.

Iu the second t r ia l . part ic ipants formed an inrpression about a

snroker rvho has problerns qui t t ing srnoking. They then judgcd

horv addictcd the target person is and then dcscribed a concrcte

disadvantage o1 smoking. The th i rd t r ia l was the cr i t ical one fbr

rvhich the descl ipt ion task perta ined to the scl f . Here. part ic ipants

f i rs t formed an impression o1'r targct person who had engaged in

cl i f fcrcnt nranual sk i l l tasks (e.g. , repair ing a bike). thcy then

. judged how rnanual ly sk i l led the target person is . and l - inal ly '

dr :scr ibed a concrete rnanua] sk i l l that they have thernselves. Sub-

sequcnt to the th i rd t r ia l part ic ipants were thankcd. dcbr ie l 'ed. and

givcn their candy.

In surn, Study 3 is based on a 2 (corrpar ison vs. coutro l p l i ln-

ing) X 2 (standard-relatccl vs. standard-ur l re latcd descr ip l ion)

mixed-nrodcl dcsign. Thc f i rs t l 'actor was manipulated between

pa( ic ipants a i ld the second wi th in part ic ipants.

I le sul ts

'l 'urget judgntent,t. On the basis o1'the rcsults ol' the prcrvious

studies, wc expecrted participants' .judgnrents about the three targt:t

persons r lot k) d i l l t r in thc cornpar ison pr iming and the contro l

group. Indeed, thcre rvas nr . r indicat ion that pr iming par l ic ipants on

comparat ive th inking led to evaluat ions of ' the target persons that

differed liorn those of control pailicipiurts. Both groups judgcd the

respect ive target persons as s i rn i lar ly happy about l iv ing in Co-

l og r re (M - 4 .28 , SD , ' . l . 40vs . M - . 4 .52 ,5 'D - 1 .03 ; / < l ) ,

tapraidS/zfr-xge/z{r-xge/2fr00109/zfr210Bd09z I xppws I S:1 I 11l2Bl0Bl 11:55 | Arl:2007-0096

CO MPA RATIVI:' THI N K IN.G l l

s im i l a r l y ac l d i c t ec l (M : ' 7 . 00 . SD = 1 .02 vs . M - 6 .91 . . !D - -

1.49: � r . : - l ) , ancl s inr i lar ly nranual ly sk i l lcd (M == 7.19, SD - 1. .12vs . M == 1 .12 . SD -ß 1 .43 ; 1< l ) . Because pa r t i c i pan t s we ree xpl ic i t ly instructecl to take their t ime in making these judgrnents,

respol lsc latcncies cclu ld not be analyzed.Sturulard dcst : r ipt ions. Our reasoninc intp l ics that pürt ic ipauts

who wcrcr proccdural ly pr imed to re ly nrore heavi ly on cornpara-t ivc th inking whcn jut lg ing the txrget pcrsou spontancously act i -vated rnorc infornration about the self as a cornparis<tn standardthan did con(rol part ic ipants. As a consequcnce. cxper i rncntalpart ic ipants should bc laster in g iv ing dcscl ipt ions that are stan-dard re lated (own manual sk i l l ) but not in g iv ing standarcl -unrc luted dcscr ipt ions (d isadvantages of l iv ing in Cloktgne. d isacl-van tngcs o f s rnok ing ) .

Again. rvc cxcludcd responsc t intes that d i f fe lcd by rnore than 2.!1fs lr'orn thc cluestion rncatrs (4.17a of all responses) liorn lurther

analysis (Bargh & Chartrand. 2000). As a couscrqucnce. latenciesl i r r cr i t ical r r rsponses were not avai lable for 6 part ic ipants, so thesubsequent analysis is basccl on thc responses ol '59 pürt ic ipr"u l ts .

Our cr i t ical dependent rncasurc is the t i rnc purt ic ipants neecled towork on thc descr ipt ion task. Rcsponse la lc:ncic:s fbr thc twostandard-unrelatcd descr ipt ion tasks (d isadvantages ol ' l iv ing in

Cologne, d isadvantages ol 'snroking) werc cornbined into onc: scoreby calculat ing the mean.

' [ 'he ureans providcd in Figure 2 are c lear ly consis lent wi th our

hvpothescs. As cxpected. part ic ipanls who werc proccdural ly

pr i rned to re ly nrorc hcavi ly on cornpar isou whcn . judging thetarget person wcre fastcr in giving thc slanclard-related description(M == 18.143 rns, , lD - . 8,262) than conlro l part ic ipants (M ==

25 ,191 rns . SD == 14 .9 ;12 ) . t ( 57 \ : 2 . 02 . p : . 02 . d = 0 .53 . I n

c:ontrast . no di f f 'erencc in the t i lne part ic ipants tcxrk to g ive thestandard-unrc latcd descl ipt ion emerged (M =- 18.169 rns. 5 '1) .=

8 .723 vs . 18 .639 ms . .SD = 9 ,250 : t < l ) . I n a 2 ( con rpa r i son vs .

conlro l pr i rn ing) X 2 (standard-rc latcd r , 's . s tandard-unrelated de-

scr ipt ion) rn ixcd-model ANOVA, th is pat tenr was borne out in as i gu i f i can t i n t c rac t i on c1 c t . F (1 .57 ) = j 9 t . p = . 05 , pa r t i a l

1t - - . { )7. In t ) r is analysis the nrain c l ' lbct o l dcscr ipt ion alsoreachccl s igni l ' icance. I ' ( l^ -57) - . 5. ( r0. p = .02. par l ia l 11r =- .119.

indicat ing that part ic ipants $ 'cre laster in g iv ing the stanclard-

unrelated thau the st l rndard-rc latcd dcscr ipt ions. The nrain ef fcct ofpr int ing did not rcach s igni l ' icance. l ' (1, 57) - 2.11. ns.

30000 r

I) i.st: Lrs s iort

T 'hese l ' indings demonstratc that colnparat ive th inking cntai ls

thrr spontancous act ivat ion of judgnrent-re latet l in fornrat ion about

thc: compar ison standard. Our reasoning holds that th is act ivat ioncontr ibutes to the ef f ic iency advantages of comparat ive th inking in

that juclgc:s lnay use the activated standard inlirrrnation as a proxy

for target informat ion that is not avaihble or is d i l l ' icu l t to obtain.The activated standard inlbrrnation nlay thus be used to lill in theblanks of rn iss ing target i l l i )nnat ion. Study 4 was designed toexamine whether such information transfcr is incleed at play ch"rringcornparat ive th inki ng.

Study 4

To do so, we again asked our part ic ipants to make judgnrcnts

about a target Ci ty X that rvas descr ibed in terrns that made i t

s inr i lar to their hometor.vn. The resul ts o1' Studies 2 and 3

dcrnonstrate that in th is s i tuat ion, part ic ipants who are pr i rned

on comparat ivc processing th ink more about thci r hor.netown

and act ivale in lbrnat ion about th is conrpar ison standard. Our

reasoning holds that th is act ivatsd informat ion about thc stan-

dard is t ransferred to the targct . f t ' th is is indecd thc case, thcnpart ic ipants pr inret l on cornparat ive th inking shclu l t l be nrore

l ikc ly than contro l part ic ipants to ascr ibc aspects that charac-

t e r i ze t he s l anda rd t o t he t a rge t . Fo r cxan rp l e , i 1 ' pu r t i c i pan t s

pr imcd on conrparat ive th inking usc thei l lxrmctown o1-Cologne

as a standard rvhen judging target Ci ty X, thcy nray ascr ibe

t yp i ca l aspec t s o f Co logne t o C i t y X , even i f t hcse aspcc t s wc rc

no1 ment ioned in the target de scr ipt ion. Tl rey rnay, for cxarnple,

a lTi rnr the possib i l i ty rhat Ci(y X has a cathedral . is a centcr of

t he rned ia i ndus t r y , and i s l oca ted on t he bauks o fa r i r e r .

Merhotl

Part ic ipants. Fi f ty- f ive studcnts at the Univers i ty of Colognepart ic ipated in thc study. They were approached in a univcrs i ty

cafeter ia, asked to take part in a ser ies o1'unrelatccl s tudies, ancl

o l1 'ered a chocolate bar as compcnsat ion.

Materittls und procedure. Participants wcre led to the exper-

imental lab and seated in indiv idual cubic les that each contained apcrsoual colnputer.

-['hc cxpcrirncnler explaincd thal participants

would work cln a serics of unrclated tasks that were adrninistered

togcthc:r sole ly tcl sarvc scarce research resources. Parl.icipants u't'r'cr

thcn given a lo ldcr that inc luded the proccdural pr iming task that

was ident ical to the one used in Study I B. Thus one group of thc

par l ic ipants 1n . - 28) wus pr imed on comparat ive processing.

rvhereas anothcr group (n : 2J) workccl on the control task.

Af ter comple t ion of the procedural pr i rn ing task, part ic ipants

cor. r t inued rv i th the judgment task, which was administcred at

t he co rnpu te r . As i n S tudy 2 , i ns t r uc t i ons exp la i ned t ha t pa r t i c -

ipants ' task \a,as to t ' i rs t form an impression and then to nrakc a

scr ies of . ludgrnents about an unknown Ci ty X. Part icrpants werc

cxposcd to thc descr ipt ion of Ci ty X, which contained a uumber

of aspecls that rnade Ci ty X s i rn i lar and thus comparable topa r t i c i pan t s ' ho rne town o l ' Cu logne . Ge rn rany . A t t he sa rne

t i rne. the palagraplr a lso pointed out that Ci ty X rvas a Scancl i -

nav ian c i t y . t hus rnak ing i t c l ea r t ha t i 1 i s no t Co logne i t se l f .

More spcci f ical ly . Ci ty X rvas descr ibed as a largcr c i ty that is

tt

o

G

o{,

U}

2 5000

20000

1 5000

10000

5000

0

i;ü;;fi;;i]

processrn9 ]

I t rcontror ]

s tanda rd -u nrelated

standard-related

Figurt 2. In Study 3. rrrean responsc lirtencies tirr standlrd-rclated vs.s landard-unrelatcd descr ipt ions by pr i rn ing (comparat ive vs. contro l ) . Emrrbars rcoLesent stanclarri crrors of the rnean.

l 2

tapraid5/zfr-xge/ztr-xge/zlr00l09lztr2108d09z I xppws I S-.1 | 11128108| ' l1:55 | Art:2007-0096

MUSSWIITI..LIR AND IlPSTUI)ll

easy t o r each by a l l means o f t r anspo r t a t i on . I t was exp la i ned

tha t i t i s a p l ace w i t h a r i ch cu l t u ra l h i s t o r y . i s an imp ( ) r t a l l t

c l c s l i na ( i on l i ) r l oL r r i s t s , and has a l i ve l y l nus i c scene . I n add i -

t i on , t he desc r i p t i on h i gh l i gh ted t he I ' a c t t ha t t hc r c i t y i s kno rvn

l i r r i ts largc uurnber of fest ivals. Sinr i lar to thc procedure of

Study 2. a l ier rc l lect ing on the doscr ipt ion and thcreby fonning

an i rnprcssion o1 Ci ty X. part ic ipants answered a ser ies o1'ques t i ons abou t t h i s c i t y . Spcc i f i ca l l y . t hey eva lua t cd C i t y X on

seven d imcns ions (nun rbe r o1 ' s t uden t s , c i t i zens , t r a i n s t a l i ons ,post of f iccs. thcatcrs, rnovie theaters. and Starbucks col ' fcc

shops ) by g i v i ng numer i c es l i n ra t cs . Fo r each es t ima tc , t he

cr i t ical qucst ion \4 'as presentcd on the cornputer scredrr (c.g. ,"How rnany un i ve r s i t y s t uden t s a re t hc re i u C i t y X ' 1 " ) . Pa r t i c -

ipants rvere asked to press the spacebar as soon as thcy had

con rc L l p w i t h an es t i n l a te anc l t o t ype i n t he i r r c sponse on t hesubscquenl scrccn. Wc rccordcd thc t i rnc that e lapsed f rom thep lesen ta t i on o l ' t hc ques t i on u l l t i l pa r t i c i pa l r t s p resscd t l l e spa -

ccbar to indicatc that thcy had forrned an cst i lnatc. Af ' ter par-

t i c i pan t s had g i ven a l l s c ' v c :n es t i u l a t cs . t hey we re i ns t r uc ted t o

wo rk on a subsequen t " vc r i i i ca t i on " t ask t ha t was des igned t c r

t 'ur lhcr explore thci r i rnpression of target Ci ty X. Thc instruc-

t i ons po in ted ou t t ha t t h i s t ask i nvo l vcd s i n rp l y gu r ' : : i ng

rvhcther aspects that are conrnron t ( ) sonlc c i r ics arc a lso prcsent

in the vagr.re ly descr ibed Ci ty X. This task involved s i rupleycs -no j udg rncn t s abou t 32 aspec t s o f C i t y X . Each aspec t was

dcsc r i bcc l by a s i ng l c r vo rd . aud pa r t i c i pan t s ' t ask was t o

indicatc 'uvhcthcr Ihe prcscnted word referred to an aspcct lhat

one n r i gh t f i nd i n C i t y X . The s t imu l i p resen t c :d i n t h i s t ask we re

sc l cc ted on t hc bas i s o f a p re tes t r cga rd i ng t he i r t yp i ca l i t y l b r

l l l e c i t . y o f Co lognc . Wc chosc t hc n l os t t yp i ca l and t he u l os t

a t yp i ca l aspcc t s c r r r r r g i ng f r o rn t h i s p r c t cs t . S i x t ecn o l ' t hep resen t cd aspec t s we rc ve ry t vp i ca l o f ' t he c i t y o f Co logne (e .g . .

l)r;in [cathedrttl]. F' c rn,s c hs e nde r l'fY stationsl, Kurnevol lcar-dr j t ' t r l l . Bruuhuas [bre wery] . Cl t r i .s lopher Streat Day). where as

l6 wcre not (c.9. , Sc/r / r . r .ss lcast le l . Ht tusl toote [ rouseboats] .l ler 'ge Inrountatnsl . 7 ' r iutnphl togen I t r iurnphal archl . EIbe l ther ivcr Elbcl . Mot lentes.se I fashion 1 'a i r ] ) . Each aspect was pre-

sentecl at the cc:nler o1' thc cornuuter sc l 'een in randotn ordcr

r vhc re i t r c rna inc t l un t i l pa r t i c i pan t s had i nd i ca ted r vhc the r i t

cha rac t c r i zes C i t y X by l - r i t t i ng t he r i gh t ( " yes " ) o r l e l t ( " no " )

tontrr t l kcv.

Af tcr conrplet ion ol ' th is ver i f icat ion task. part ic ipants ucrc

thrrnked l i r r thc: i r part ic ipat ion, debr iefed. : rnd of fcrccl their corn-pc l l sa t r on .

Rc.rrrll.s

'litrget jutlgntent.s. We expectcd to replicate thrr fcsults cr1'

Studl '2 and s l rorv (hal part ic ipants who wsrc pr inrcd on compar-

at ive th inking arc fastcr than contro l part ic ipants in judging Ci ty X.

To analyzt: rcsponse latencies lbr the critical targc:1 .judgments.u,c excluded response t i rnes (hat d i l ' f 'ered by more than 2. ! / )s f rom

thc c luest ion nrean (3.90% of a l l responses) f rorn fur ther analysis(Bargh & Chartranr l . 2000) and avelagcd the rcrnain ing Iatencies

into one int lex (Crorrbach's u - .69). As expectc:d. paf t ic ipant \

rvho wcrc ; r r imed on conrparat ivc th inki r rg wcre f aster in evaluat-

ing Ci ty X (M : 6. .591 rns, SD : 1,994) than contro l part ic ipants(M ' 8.048 ms. S/)) - 2. ;184), r (53) - 2.41, p '= .02. d - 0.54.

In addi t ion. rvc again exarr incd rvhether part ic ipants in both

concl i t ions judged target Cl i ty X di l lcrent ly . To do so. we cxcludecl

est imates that c leviatccl f ronr the quest ion mean by nrol 'e than 2

SI)s, r - t ranslorrned the indiv idual judgnrcnts, ancl averaced thern

into one index. Part ic ipants ' . judgrnents o1' target Ci ty X did not

dc:pend on rvhether thcy wcrc prirned on conrparative processing(M = 0.06-s. SD : 0.62) or not (M : -0.02. SD : 0.55; t < l ) .

Asped verificalions. Wc hypothesizcd that participants who

wc:rc pr i rned ou colnpalat ive th inking would bc rrore l ikc ly to

ascribe aspects that characterizc the stilndard C)ologne to the target

Ci ly X. Part ic ipants who were pr imed on cornparat ive th inking

should thus be morc likely than conffol participants to acccpt

aspects lhat a|e typical ( ) { Cologne as charactcr iz ing Ci ty X. For

aspects that are not typical ol'Cologne. tlre acccptance rates should

not dill 'er firr both groups.

To examinc th is possib i l i ty , we f i rs t calculated two sunl

scorcs: oue represent ing thc nurnber of accepted xspccts that are

relatcd to the standard and one represent ing lhe nurnbcr of '

accepted aspects that are not rc lated to the standard. As inspec-

t ion ol F igure 3 rcvcals, the acceptance ratcs in the ver i f icat ion

task were consistent wi th onr expectat ions. Indeed, pal t ic ipants

pr imed on compal 'at ive th inking acccptcd more C- 'o lognc-relatecl

st inrul i as typical for Ci ty X (M : I 1.86, SD : 1.14) than did

con t ro l pa r t i c i pan t s (M : 10 .18 . SD = t . 39 t , ( 53 ) : 3 . 98 .p -

.001. c i : 1.07. No such di l - ference was fotrnd for Colognc-

unrelatet l aspects (M = 6.-53. SD : 2.62 vs. M : 6.56, SD --

2 .66 : t < l ) . Th i s pa t t e rn o f n l c : ans resu l t c ( l i n a s i gn i l i can t

two -way i n t c rac t i on i n u 2 ( comp l r i son vs . con t ro l p r i r n i ng ) X

2 ( s t anda r t l - r e l a l ed vs . s t anda rd -un re l a ted aspec t s ) r n i xec l -

moc l c l ANOVA, f ( 1 . 53 ) : J . 14 . f < . 0 . 5 . pa r t i a l 12 = . 07 . I n

th is analysis, the main ef fecl for the type of aspect was also

s i g l i f i c a n t . f ' ( 1 . 5 3 ) - 1 1 8 . 5 3 , p : . 0 0 1 . p a r t i a l 1 1 r = . 6 9 .

Overal l . part ic ipants were morc l ikc ly to accepI standart l -

re lated aspecls (M = | 1.04. SD - 1.70) than standart l -unrelatet l

aspec t s (M - 6 .5 -5 , SD - 2 .62 ) . I n add i t i on . t he ma in e f l ec t 1b r

p r im ing was s i gn i f i can t , t ' ( 1 , 53 ) = ' 3 . 90 . p == . 05 . pa f t i a l I r -

.069. suggest ing that part ic ipants in the cornparat ive th inking

condi t icrn tendccl to accept Inore aspeLrts (M :9.20, SD - 2.21)

t hun d i d con l r o l pa r t i c i pa r t s (M : 8 .37 , . tD - - 2 . 15 ) .

rl

oLru3 1 4or ! ' -

o . ^> r u

G o

L

o

o

z

li;";t;;;if;;lI

processrngi l - l r n n t r n l

L - ' " - " 1

Figure 3. ln Study 4. rrurnber of standard-rclated vs. stand;ud-unrelltedaspccts that rvcrc acceptcd as charactcrizirrg the largct by prirning (cotn-parativc vs. col.ltrol). Ilrror bars rcprescnt standard errors of the nrean.

sta nda rd-related sta nd a rd -u n re lated

t ap ra id5 /z f r - xge /z f r - xge /2 f r00109 /z f r2 t08d09z l xppws lS .=1 l t 1 t2B t }B l i 1 :55 lA r t : 2007 -0096

CONIPA RATIVIJ TFTINK tNG

I) isc tt,s.sio tt

Thesc f i nd i ngs have a t l eas l two no tewo r thy i r np l i ca t i ons .First and l i r re most. lhey demonstrate that colnparat ive th inkinginduccs. judges to ascr ibe aspects that characler ize the judgnen-

ta l s t anda rd [ o t he j udgmcn ta ] t a rge t . I l l l he con t cx t o f t h i sstudy. part ic ipants who wcre proccdural ly pr i rned to cngagr ' innrore conrparat iVe th inkinc wcre ntorc l ikely than conlro l par-

t r c i pan t s t o i nd i ca t c : t ha t aspec t s t ha t r r e t yp i ca l o l ' t he s tanda rdCo lognc a l so cha rac te r i z c t hc t a rgc t C i t y X . [ n l i ne w i t h t hcp resen t r eason ing . t h i s dc rnons t ra t cs t ha t access ib l c s t i l nda rdir r furmat ion is inc leecl uscd as a proxy lbr target i i l l 'ornrat ion thati s no t ava i l ab l c . Th i s f i nd i ng l u r t hc : r ' spec i f i e s l r ow c r rn tpa ra t i v r -th inking shapes in l i r rn ' rat ion processing. Study 2 demol ls(ratedthat part ic ipants who are pr i r rcd on comparat ive th inking th inkrnore about a pcl t incrnt standard dur ing target evaluat ion. Study3 showed that doing so rrakes infonnat ion ahout thc ccrnrpar i -son s tanda rd acccss ib l e . S tudy 4 sugges t s r ha t t h i s acccss ib l estandard in lbrrnal ion is indeed t ranslcrred to lhe target .

In addi t ion, the prcscnt t ' inc l ing also repl icates thc judgntetr t

fac i l i tat ion el1ect f i rs t obtainc:d in Stucly 2. In l ine wi lh ourana l ys i s , wc aga in show tha t pa r t i c i pan t s p r i r ncd on co lnpa ra -t i v c t h i nk i ng a re l ' r s t c l i r r n rak i ng t h r : c r i t i ca l t a rge t j uc l gmen ts

than con t l o l pa r t i c i pa r r t s . As i n S tud i cs 2 and 3 . t hc re i s noind i ca t i on t ha t t hc j udg rnen ta l ou t con t cs i n bo th g roups d i f -l ' e r cd . ' l ' h i s aga in sugges t s t ha t con )pa ra t i v c t h i nk i ng ho ldscl f ic icncy advantal lcs. Study 5 was dcsigne<l to fur ther substan-t i a t c t h i s poss ib i l i t y .

Study 5

So l - r r , rve have lested lor the assurned ef f : ic icncy advantagcso f co rnpa ra t i ve i n l o rma t i on p rocess i r r g by cxam in ing d i l l e r -cnccs in thc t i rne. judges nc:cd to ntake a cr i t ical target . judgrnent.We hypo thes i zcd t ha t con rpa l l t i v c p roccs : i r r g u r r u l d s imp l i l ' yt hc j udgn ren t p roccss and save cogn i t i ve capac i t y . I f t h i s i s t hecase . I l l cD pa r t r c i pan t s shou ld he ab le t o n rakc a c r i t i ca l . j udg -mcn t l - a \ t c r i I t hcv r e l y mo re hcaV i l y on con ipa r i s< tns . S tud ies 2and 4 c lcnronstratccl that th is is indeed the case. Furtherrnore. i fcompa ra t iVe p roccss ing does i ndc : cd save cogn i t i ve capac i t i c s .then judges u,ho re ly nrore hcavi ly on cornpar isons in rnaking atargct judgnrelr t should rr rquire lcss processing capaci ty t0 d()so . As a consequence , mo re r cs i dua l capac i t y shou ld bc ava i l -ab l c f r t r a secouda ry t ask on wh i ch judges r vo r k i n pa ra l l e l . Jus tas ca t cgo r i ca l t h i nk i ng , f o r cxa rnp le , f r ecs cogn i t i ve capac i l i c sfo r a pa ra l l c l t ask ( e .g . . Mac rae c t a l . . 1994 ) , co rnpa ru l i veth i nk i ng shou ld l ' r ce capac i t i es t ha t can t hcn be used t b r aseconda rv t ask . I n S tudy 5 , wc uscd a c l ua l - t ask pa lad igm(Ba rgh . l 9 f J2 : l \ 4ac rae e t a l . , 1994 : Ne rvon & Gophe r . 1979 :W ickcns . 1976 ) t o exam ine t h i s poss ib i l i t y and t o t hus shedlu r t hc r I i gh t on t he po t cn t i a l c f f i c i ency ac l van tages o f co r rpa r -a t i v c p rocc r ss i ng . Wh i l c f o r r n i ng t hc i r i n t p ress i on o f 'C i t y X ,pa r t i c i pan rs s i n ru l t ancous l y n ron i t o red a t ape - reco rded t ex t . l fconrparat ivd proccssing docs indced savc: cogni l ivc rcsourccs.l hc r r pa r t i c i pan t s r vho re l y r nu re heav i l y on compar i sons l vhenl i r rming an i rnpression ot ' the targct should havc rnorc resic lualcapac i t y ava i l ab l c 1o mon i t o r t he t ex t . As i n p rev i ous r csea rch

using a s imi lur procc:dure to exanr inc the ef ' l ' ic icncy advantageso1 ' s t e r co t yp ing (Mac rae e t a l . , 1994 ) , t h i s shou ld a l l o i v pa r t i c -

ipants to t i rcus more on the detai ls of the storysubsequcnt ly ablc 1o reproduce rnore of ' these

l - 1

so tha( thev are

detai l s .

Metltod

I'articipttttts. Filiy-iive studcnts look part in the study. Theywcrc approached in the university cal'etcria and asked to partici-pate in a psychological exper intent in exchange for a chocolate bar.

ltrocedure. Upon agreentcnt, participants rvcre lcd to a sepa-rate roonr and seated in indiv idual cubic les. Mater ia ls and procc-

dures were s imi lar to those used in Str"rd ies I and 2. Using the sanremater ia ls as in Studies lA and 2A, pf f t ic ipants were pr inred toprocess inlirrnration in either a rnore or a less cclntparatlve manner.The seconcl part o l ' the study was s inr i lar to the judgment lask usedin Stud), 2. This part was introduced as au exper i rncnt on thecf fects of mental d ist ract ion on a. i r . rdgmenl task. Part ic ipants w'ereinstnrctcd to put on hcadphones. They i i rs t formed an inrplcssiou

ol ' the target Ci tv X f rorn the sarne wr i t ten descr ipt ion used in

Study 2. Al'ter lbrming this inrpression. the experilnenler startcd

the prcsentation ol an audiotaped story, which lasted fbr a total o1'

about 3 min. The story was about a wornal l who had t rouble wi th

her best l r iend because o1'a canary. Par l ic ipants were instructed tosi rnul taneously rnoni tor the detai ls o l ' th is story whi le judging thetarget Ci ty X on the same dinrensions used in Study 2. Part ic ipants

thus evaluated Ci ty X whi lc l is tening fo thc tapc. Af ter cornpl t : t ion

of the judgment task, part ic ipants were askcd to wr i tc dorvn as

mnny dcrlails from the short story as thev could rcrcall.

Results' l -arget

. ju lgrnants. On 1hc basis o1' the resul ts o{ 'Studies 2 and

4, we again expected part ic ipants rvho rvere pr i lned on a morccompalat ive processing nrode to be faster in judging the targel thal)

contro l part ic ipants. To cxamine th is possib i l i ty , we aeaiu c:x-cluded rc:sponse times that dilfcred by more than 2 SDs fton the

rnean (2.97c ol a l l responses). As belorc, thrr mLran responsc l inrefor a l l t ivc judgments on Ci ty X was conrpl l ted fur each part ic i l ) i ln t(Cronbuch's a =. .65). Resul ts were in l ine rv i th those of Stucl ies

2A and 28. Part ic ipants in thc conrparat ive processi t rg condi t ionntade the critical judgments läster (M --. 6,562 ms, .9D - 2.024)tlran did control particip ants (M .. 8. 1 30 nrs. .11) =. 2,632), /( 53) --

2.47. p : .01 , d :0.67. Thus, cornparat ivc processing faci l i tates

the cr i t ical target judgnlcnts even i f . ludges s i rntr l tancouslv work ona sc:contl task rcquiring attentionäl resourccrs.

Once agairr , rvc cornpared the actual judgrnc:nts part ic ipunls

made fbr City X. We exch.rded estimates that dcviated from thequesti(nr rncan by nrore than 2 .5'Ds, :-translbrmed the individualjudgrnents, and averaged thcru into onc index. Part ic ipants ' . juc lg-

ments ol' target City X did not dc:pend on whether thev wcrepr i rncd on cornparat ivc: proccssing (M - -0.1 l ,5D .= 0.48) or not(M = 0.12, SD = 0. ,s9). t ( -53) : 1.59. n.y.

Serr tndurt task. Thc rnore cr i t ical measure in Study 5, how-

ever, is participants' perlonnance in thtr secondary task, namely.

thr: numbcr of rcrncrnbered details front the short story. 11'com-par l t ivc pnrcessing saves cogni l ive resoulcL:s. thcn part ic ipants in

thc compar ison condi( ion should perfurm bct ter than those in thecontro l condi t i ru. To exarnine th is possib i l i ty . we counted thenuurber of correct ly reported detai ls for each part ic ipant . Consis-tcnt wi th or , l r cxpectat ion, part ic ipants who wcrc pr imed on a nr i l re

t 4

tapra id5/zf r -xge/zf r -xge/zf r00109lzI r210$d09z lxppws lS=1 111128108 |111:55 lAr t : 2007-0096

MUSSWI]I I -L]R AND I ]PSTUDI]

cornparat ive pnrcessing mocle renrcnrbered morc detai ls (M -

,1.96. . tD - .= 0.91) than c l ic l contro l part ic ipants (&/ : ?.19, SD =

0. .54). / (52) '= 2.0!) . t , : .02. r1 -= 0. .5-5.s

l)l.i<'tts,iiorr

These f indings repl icate and extend those o1'Studies 2 and 4 ininrpor lant ways. Fi rst . they again deuronstrate that rc ly ing rnoreheavi l -v on conrpar isc lns al lows . judgcs to ntake a cr i t ical target

. judgnrcnt fastcr . I t is notable that th is faci l i ta l ion ef fcct is apparent

evcn i l 'par( ic ipants concomitant lv u,ork on a sccond task that

dernands addi t ional proccssing capaci t ies. Furthermorc, Study 5

dcrnonslrates the ef ' f ic iency advantages of cornl tarat ivc in l t rnna-

t ion processing on an ent i rc ly novel task. Not only does compar-at ivc processing specd up thc cr i t ical . judgrncnt i tsel l ' , i t a lso l ieescogni t ivc capaci t ies that may thcn be used tu rvork tx a secondary

task. Becausc part ic ipants could a l locate inorc at tcnt ional re-

sourccs to thc shor( s1ory. t l lcy wc:rc bet ter ablc Lo recal l i ts dela i ls .Again. there is no indicat ion thal corrparat ive th inking changed thejudgrncntal oulcornc, suggest in{ : that the increasc in c l f ic iencylnay not come at thc cost of a decrease in accuracy.

Stucly 6

Arc thcrc l i rn i ts t< l thc: cf f ic i t :ncy advantages o1' ct rmparat ivc

th inking that havc been showu in thc prcvious studics 'J Our rca-soning suggcsts a l ' i ls t potent ia l boundary condi t ion that inf luenceslhc extent to which thcsc c l l ' ic iency advantagcs wi l l be apparcnt . I f

cot t rpur isons arc carr ied out a long an al ignable structure thatre lates indiv idual fcatules of targct and standard to one anothef ,then t l . rc case wi th u 'h ich such an al ignable structure can beestabl ishccl should c leternr ine thc lnagni tude of the obtained el l ' i -c icncy rdvantagc. Study 6 was designed to exaln ine th is possib i l -

it1'. To do so, wc adapled a product choice task that has bccn

dcvclopcd by Zhang and Marknran (200 1 ) . Part ic ipants werc askcd

to colnpare two typcs ol 'popcorn descr ibed by four at t r ibutes each

and to i r rd icate which onc thc1, prc l t r . For one hal f of 'our part ic-

ipants a l l l i rur af t r ibutes wcre nonal ignable in that they conccrnc( l

d i f fcrent aspects that are re lat ively indepcndent of onc another(e.g. . prcparat ion in i ts own bag vs. tastcs sclmewhat swcet) . For

thc othcr hal1. two ol the at t r ibutes wcrc a l ignable in that thcy

conccrnr:c l a s i rn i lar aspect ( r : .g. . preparat ion in i ts own bag vs.

preparat io i l in a microwave borv l ) . For thc: l ' i rs t group i t is d i f f icul t

to establish an alignahle structure and to thus compilre both prod-

ucts. For the sccond group i t is easier to estahl ish an al ignable

structure to comparc both products. We hypothesizcd that thc:

c l f ic iency a( lvantages of cornparat ive th inking would l rc more

pronounced i l 'an al ignable structulc is casicr to cstabl ish.

Metlrttl

Purt i t iput t ts . Wc rccrr . r i ted 55 students at the Univers i ty of

Colognc as part ic ipants. Thcy rvere approached in a univers i ty

calcter ia, askcd to take part in a ser i r :s of uurc lated studic:s, ando f I r ' r c t ] a e l t r r c r r l : r t c h l r l r c r r l n I cn : : t t i o t t .

Mut?t i ( r l . t . The mater ia ls wc usccl lor the product choicc task

wcrc adaptcd l l 'onr thosc c lesigncd by Zhang ancl lv larkrnan (2001).

Part ic ipanls rvcrc askcd to decide rvhich ol ' two brands of popcorn

they prelcr . I lach brand rvas crharactcr ized hy four at t r ibutcs. Two

of these at t r ibutes were nonal ignable in thal lhey perta ined todist inct d imensions l i r r both brands ( large kenrels vs. s l ig l r t ly low

in corn ancl grain l'lavor; crunchy for a long time vs. waterproof

wrapping). For part ic ipants in the nonal ignable condi t ion. therenrain ing at t r ibutcs were s imi lar ly nonal ignable and also per-

ta ined to d ist inct d imensions (preparat ion in i ts own bag vs. lastes

a bi t sweet l calor ies eqr. ra l to a s l icc o i bread vs. has some ci t r icacid) . For part ic ipants in the al ignablc condi t ion. thcse two rcrnain-

ing at t r ibutes were al ignable in that they pcrta ined to the salne

dirnensi t>n for both brands (prcparat ion in i ts own bag vs. prepa-

rat ion in microwave bowl; calor ies equal to a s l ice o1 brcad vs.

calor ics equal to a tablespoon of ' sugar) . This in lbrmat ion was

givcn to part ic ipants in the form of a table (see Table l ) .

Prot:atlures. Upon agreement, participants rvere lecl to a sep-

arate lab roorn and seated in indiv idual cubic les equipped wi th

personal colnputers. The experinrentcr first explained that partic-

ipants would work on a series ol' scparate studies that were cou-

ducled togc:thcr to save scarce research resources. Participants

were instructed to firsl wolk tlrrough the rnaterials in a fblder that

was located in front of them. This lblder inclr"uled thc proccclural

pr inr ing task, which was ic lent ical to the one used in Study 3. Thus,

al l pal t ic ipants wcre again exposccl to the pictnre of ' the junglc

scene. A first group ol'participants fu - 2'7) w'as asked to conlpare

the left and right halves of the picture and to write down all thcr

s in i i lar i t ics and di l ' lerenccs they could l ind. Clontro l part ic ip i lnts(n - 28) were asked to describe thc: picturc:.

Upon complet ion of the procedural pr iming task. part ic ip lnts

rvcre told to procced with thc next task. rvhich was computer

adnl in istered. lnstruct ions pointed out that lh is study was con-

cerned wi th product cvaluat ions. More speci l ' ica l ly . part ic ipants

rcccir r -d thc l i r l lowing instrucl ions:

I rnagine you would l ike to huv nr icrowave popcorn. In the store youhavc the choice hetwcen tu'o brancls. Please tbrm an impression ofboth brunds on the basis of thc information yoLr rvill reccive, We rvilllsk you a few question about thcse popconrs. To answcr "Popconr A"pleascr press the hutton markcd with a blue sticker with your left inclexfinger. To answcr "Popcurn B" plcasc press the button marked with ayel low st icker rv i th youl r ight indcx l inger ' . Please put your indextingers orr thc rcspective keys.

Bekrrc bcing cxposct l to the product informat ion. part ic iprnts

workc:tl on four practice trials that wcre included to farniliarizc

them with thc rcsponse procedure. 'Io

practice. participants wl-rr-

asked to ansu,er "Popcorn A" and "Popconr B" trvice each. At the

end of the four pract icc t r ia ls, part ic ipants werc in lorrncd that lhey

would now be given thc cr i t ical infbrmat ion about the brands of

popconr. Thcy wcrc remindcd to put their index l ingers on thc

respect ivc kcys and to procecd to thcr next screen by prcssing the

spacebar.

This screen contaiued thc: infornration about the two brands o1'

popcorn. prcsented in a tablc (see' l 'able l ) . T 'he c l i t ical quest ion."Which popcorn would you preter'?" was presented on the top ot'

thc screen. and thc choicc options "Popcorn A" and "Popcurn B"

were prescntcd on the lef i ant l r ight s ic les of ' thc bot torn of the

scrcetr , respect ivc ly. This in l r l rnrat ion renrained on the screen unt i l

'Orrc part ic ipant t i i lcd to answer the recal l quest ion; thcrefore, th isanalvsis is hasct l on thc responses ol 5:1 part ic ipants.

'n

tapraid5/zfrxqe/zfr-xge/zfr00109lztr2108d09z I xppws I S:1 | 11128108 | 1'l :55 | Art: 2007-0096

COMPA RATIVII TFt ln\ K ING t 5

Preparation in ifs ou'n bagLargc kernc lsCrunchv fo r r )ong t i rne

Tasles a hi t sweetSl ight ly lou ' in corn aud grain f lavorWaterproof wrapping

Tablc I

Infrtrnrutiort About '['u,o

I]rtuttls o.f Popcorn Proyüled to

I'urtitipurrts in tItt NonuIignahIa t,s. AIignultle ()orttlitiort

(Srut l t , 6)

Popcorn A Popcorn I3

Nona l ignab le cond i t ion

.03, part ia l l : - .09. In th is analysis. none ol ' thc retnain ing ef ' t 'ects

reached s i gn i f i cancc (a l l Fs < l ) .

I t is notable that part ic ipants ' choices between Popconr A ancl

Popcom B rvere nnt s igni f icant ly in l luencecl by our exper imental

nranipulat ions. In a 2 (corrpar isorr vs. contro l pr inr ing) x 2 (h igh

vs. low al ignabi l i ty) between-subjects ANOVA using part ic ipants '

choiccs as the dcpendenl lneasure, nei ther the main ef fect of

a l i gnab i l i t y ( I - < l ) no r t he ma in c l l t c t o f p r im ing , f ' ( 1 . 5 l ) =

2.09. p = .15, nor the interact ion ( f '< l ) reachcd s igni f icancc.

D i scu ss i o t t

These f indings havc a( least two notable impl icat ions. For txc.

they cstabl ish a f i rs t boundary concl i t iun ol ' the elTic iency advan-

tagcs of comparat ivc th inking. Part ic ipants rvho rvere pr inred on

conlparative proccssing were faster to choose which popcorn thcy

prcl'ered only if the features of both products were at lcast partly

al ignable. F 'ur thernrure. these f indings indirect ly support the not ion

that cornparisons are carried out along alignable structures that

interconnect I€atures of target and standard and thus echo the

impl icat ions of a largc: body of evidence (1or an overv icw, see e.g. ,

Gentner & Markrnan. 1997). l f s t ructural a l igntnent is i rnpeded so

that conrparisons arc more dilficult to carry out, then the elTiciency

advanlages of comparative thinking fade. ln this respect. struclural

a l ig lnrent appcars to conüibute 1o the el l ic iency of cotnparat ivc:

th i r tk ing.

Srudy 7

We designed oLrr las l s tudy to exanr ine a seconcl potcnt ia l

boundary condi t ion, nanrely. the evaluabi l i ty 01'at t r ibutes. I t has

becn suggested that attributcs cliffer in the cxtent to whiclt lhey carl

be evaluated in isolat ion versus require a conrpar ison (Hscc: , I 996).

Whcthcr a house that one considers buying does or docs not have

a garden, for cxample. can bc easily c:valuated tvithout comparing

i t to other houses. Whether the pr ice f r r r the house is rcasonable.

lrorvever, depends on a corlplex array of additional läctors (e.g..

location, market priccs in thc: iu'ea) and is thus dilficult to evaluatc

wi thout cornpar ing i t to a l tcrnat ives. We hypothesizc:c l tha( the

cf f ic iency advantagcs ol 'conrparat ive th inking are pr imar i ly ap-

parent in choice s i tuat ions that involve at t r ibutes that are di l l ' icu l t

to cvaluate indepc:ndcnt ly . ' lo exarnine th is possib i l i ty . we adapted

Calor ies coua l to a s l i cc o f b rcad Conta ins c i t r i c ac id

A l ignab lc conc l i t ion

Prcparation irr its orvn [ra,g Prcprratior in microwave bowl[ -a rgc kerne ls Couta ins c i t r i c ac idCi runchy l i r r a long t iu re Waterproo f rv lapp ing

Ci r lo r ics e r lua l to a s l i ce o1 'b rca i l Ca lL r r ies equa l to tab lespoon o f sugar

part ic ipants nrade their choice. We recorded part ic ipants ' process-

i rg t inre f rorn the prescntat ion of the procluct infbrnrat ion unt i l

thcy incl icated thci r choicc.

Subscquent ly, part ic ipants rverc thanked Ibr their part ic ipat ion.

dchr ic lct j . ln t l o l ' l i ' rcr l ths i r corntrcnr l r t ion.

lle s u lts

' [ 'hc prcsent analysis holds that the c l f ic icncy advantages of '

comparat ive th inking ale rnore pronounced i f an al ignable struc-

turc belween the choice al ternat ives is casier to cstabl ish. This

should be the casc: i f at leust sornc: of their at t r ibutcs arc a l ignable

because t l . r r :y pcr ta in to thc sarnc dimension. In th is s i tur t ion,

part ic ipants pr inred on cornparat ivc th inking should be faster in

nraking thci r choicc thar. r conlro l part ic ipants. I t ' the choicc aher-

nat ives ard character izcd onlv by at t r ibutes that perta in to d ist inct

d inrcnsions. thcn thL: a l ignablc s l ruclurc a long whicrh the conrpar-

ison is carr icd out is d i t l icul t to establ ish, so thc ef f ic iency advan-

tages ol 'conlparat ive th inking may facle. In th is s i tuat ion. p i r r t ic i -

pants pr i rnccl on cornparat ive th inking s l rould not bc laster in

nraking t l re i r choice than conlr-o l part ic ipants.

Again. we cxcludcd response t i r l res that d i f fered by nxrre than 2

5' / )s f ronr thc quest ion nreans ( 1.97o o1'a l l responses) f ronr fur ther

analysis (Balgh & Chart land. 2000). As a consequencc. la(encies

for cr i t ical responscs werc not avai lable lor I part ic ipant , and thus

thc analysis of rcsponse latencics was cr)nducted on 54 part ic i -

l )an r s.

The rncan response latencies for part ic ipants ' choicc, depictc:d in

Figurc 4. arc consistent wi th our reasoning. As expectcd. rvhcn the

provided infornrat ion wrs at lcasL part ly a l ignable, part ic ipant \

rvho wcre pr inred on c i lnrparat ivc processing chosc belween the

brands ol' popcorn more cluickly (M : 16,681 rns. S/) = 5.566)

than did contro l part ic ipants (M : 20.803 nrs. 5 '1)) = 5.250).

t ( 50 ) = 1 .81 .1 = . 04 . d : 0 . , 19 . Whcn t hc p rov i ded i n l b r rna t i on

\ \ ,as not a l ignablc. howevcr ' , part ic ipants pr i rncd on crr tnparat ive

thinking wdrc not fastc:r in r raking their choice ( ,4, / - 20.639 ms,

SD = 7,28 l ) than contro l part icrpants ( lV , , , . 11 ,415 rns. S/ . ) =

5. .1 l 3) , / (50) - . 1 .39. ns. This pat tern produced a s igni f ic^an(

intcract ion cf f tc t in a 2 (compar ison vs. Lx)ntro l pr inr ing) x 2 (h igh

vs. lo iv a l ignabi l i ty . ) betu,een-subiccts ANOVA using part ic ipants '

cho i cc l a t enc i cs r s t he dcpcndcn t l neasu l ' c . F (1 .50 ) : 5 . 10 .p :

a l i g na b l e non -a l i g nab le

Figure 4. In Study 6. nrean rcsponse latcncies for the choicc betwcen trvtr

b rands o f popcorn by a l ignab i l i t y (h igh vs . low) a ld p r i tn ing (c r t tnpara t ivc

vs. control). I lrror bars reprcscnt standard cnors of the mean.

o

o.9g

(uG

(,o

oood

25000

20000

1 5000

10000

5000

0

l 6

tapraid5/zfr-xge/zfr-xge/zfi l0109lzttz1}Bdlgz I xppws lS==1 | 11/28/08 | 11:55 lArt:2007-0096

MUSSwI: l l - l iR AND I IPSTUDII

a product choise lask that has been devcloped by Hscc (1996).

Parlicipants rvere askccl to comparc: 1wo CD changcrs and choose

which onc thcy prel'er. Both products were clescribcd with respect

lo four character is t ics. One hal l ' <t1 ' our part ic ipants was givcn

background infbrmat ion about thesc character is t ics so that they

wcrc in a posi t ion to evaluatc: each product indcpendeut ly (h ig l t

evaluabi l i ty) . The sccond hal f rvas not g ivcn th is background

infornrat ion and could cvalr ratc only the i rnpl icat ions 01'sonre 01'

the chiuxctcr is t ics by cornpar ing both products ( low cvaluabi l i ty) .

Wc expccted the el f ic iency aclvantages of contparat ivc th inking to

be pr i rnar i l l ' apparcnt under condi t i r rns of l r rw evaluabi l i t l , ' .

Methrxl

Part i t ; iput t t : ; . Wc recrui tcd I l0 Univcrs i ty of Colognc stu-

dents as pal ' t ic ipants. ' l 'hcy

wcrc asked to take part in a ser ics t t f

unrelated studies and of l 'ered a chocolatc bar as compcnsrt ion.

Mutct'iuls. 'l '[re

niatcrials rvc used firr the product choice task

rvcre adapted t i 'orn those introduced by l lsee (1996, Study 4) .

Part ic ipants werc askcd to dccidc rvhich of t rvo CD cl rangers lhey

1'2,4Q:r would prc: f r : r ancl wcre givcn thc in lornral ion provided in Table 2.

The CD changcrs were charactcr ized akrng four d imensions. Two

of thosc (brand. CD capaci ty) wcre high in evaluabi l i ty in that their

inrpl icat ions cor. r ld easi ly be cvaluatcd in isolat ion. wi thout a

conrpar ison rv i th an al temat ive. ' l -he renrain ing lwo ((o la] hartnonic

distor l ion [ ' l ' ] {Dl . rvarranty) u,ere low in evaluabi l i ty in that their

inrpl icat ions could hald ly be cvaluatcd in isolat ion. unless addi-

t ional in lb l rnat ion about thci r nreaning rvas provided. This addi-

t ional infornrat ion was given to part ic ipants in the high evaluabi l -

i ty concl i t ion. Spcci l ical ly . these part ic ipants were in l i r r rned that" for n l rst products on the nlarket THD rangcs f rom 0.002%, (best)

to0.()12(/ t (worst)" and that " infornt i t t ion on warral l ty was gi l 'cn in

nlonths." ' l 'h is

addi t ional inf i l ' rnat ion al lorved part ic ipants to cval-

uate thcsc at t r ibutes for each ()1 ' thc choice opt ions wi thout coln-

par ing i t to the i r l ternat ive. ' [ 'hus.

part ic ipants in the high evalu-

ab i l i t y conc l i t i on i . ve r c : ah l c t o eva lua te each o1 ' t he g i ven

dimcnsions u ' i thont cornpar ison. Part ic ipants in thc lor .v cvaluabi l -

i ty condi t ion r .verc able to cvaluale hal f 'o l ' the given c l i rnc: t ts ions

wit l . rout corr . rpar ison.

ProLulures. Upon agrccnrcnt . part ic ipants wcrc led to a sep-

aratc lab roonr ancl scated in indiv idual cubic lcs equippcd wi th

personal cornputcrs. The exper intentcr f i rs t expla incd that part ic-

ipants rvere about to wrirk ou a series o1'separate stuclies that werc

conducted togcthcr sole ly to save scarcc research resot t rces. Par-

Tab l c 2

I rtJbrn rctli t trt Alutut C D Cltattgers I) rov i tlt 'tl to Iturlit' iltttnls(S tud t 7 )

V r r i r rb lc C'l) Change r r\ ('l) C hungr'r B

t ic ipants were instructed to l i rs t work through the nrater ia ls in a

l i r ldcr that was located in l ront of them. This lo lder inc luded thc:

procedural prirling task. rvhich was identical to the onc used in

Study 3. ' Ihus, again al l part ic ipants were exposed t t r the picture of

thc jungle scene. One group of thc part ic ipants 0t = 56) was asked

to conrparc the lef t and r ight halves of ' thcr p icture and to wr i tc

dorvn al l thc: s i rn i lar i t ies and di f fcrences they could f ind. Control

participants 0z = 54) werc asked to describc: the picturc.

Upon complet ion o1 the procedural pr i rn ing task, prut ic ipünt \

were to ld to proceed wi th the next (ask. wl t ich was col l rputer

adnl in istered. The nrater ia ls and procedures o{ ' th is task c losely

fo l lowcd those o1 Hsce (1996). lnstruct ions explained that th is

stucly rvas concerned wi th product evaluat ions. Part ic ipants rvcrc

askc t l t o i n rag inc t hc f b l l ow ing : ccna r i o :

You have decided to buy a CI) changer. In the storD. you have to

rcalize that the shop assistan( is unable to provide you with intbrrna-

lion abotrt the products that goes bcyond what is provided in the brief

product description. Your choice has narrowed down to two products

and you havc sot t te intbrrnat ion avai lable to help you wi th your

tlec ision.

Part ic ipants in the lorv cvaluabi l i ty ccrndi t ion were fur ther in-

formed that in the provided infonnation 'l ' l-lD

stands tbl total

harrnonic d istor t ion, for which a low value indicatcs bet tcr sound

clual i ty . Part ic ipants in the high evaluabi l i ty cortd i t ion rcceived the

sarne inlbrrnation and were additionally informed that for most

products on the markct Tl{D tangc:s liotn 0.0(}27c (best) to 0.012c/(;

(worst) . These part ic : ipants were {ur ther informed t l la t the warranty

was givcn in rnonths.

Bcforc being exposed to thc product in lbrnrat ion. part ic ipants

worked on four practice trials that were includcd to fanriliarizc

thcnr wi th provid ing answers using the col t lpuler keyboard. In-

struct ions to lhese pract ice t r ia ls f i rs t inkrrrncd part ic ipan(s that wc

woulc l ask thern a nunrber of 'quest ions about the CD changcrs. To

answcr ''CD

Changcrr A" they should press thc kc:y rnarked with a

blue st icker wi th their lef t index l ' ingcrr . ' l 'o

answer "CD Changer

8," thcy should press thc kcy tnarked wi th a ycl lorv st icker wi th

thci r r ight index f inger. Part ic ipants wcrc to ld to keep their indcx

{'ingcrs on the respective keys throughout the 1ask. To practice,

participanls rvcrc asked to answer "CD

Changer A" and "CD

Changer B" twicc each. At the cnd of the four pract ice t r ia ls,

part ic ipauts were iufornred that they would nolv be given the

cr i t ical in lormat ion about the CD changers and asked lo eval-

uatc thLrm. They wcre renr inded to put their indcx l ' ingcls on t l te

respectivs keys and to procecd to the next scrcetr by prcssing thc:

spacrcbar.

This screen contained lhe informal ion about the two CD chang-

c:rs, which was presentcd in a table s i tn i lar to ' I tb le 2. Thc cr i t ical

question, "Which CD changcr u'ould you prefer?" was prescnted at

thc top of the screen, and the choice options "CD Changcr A" and"CD Changer B" wcrc prcsented on the lc l i and r ight s ides of rhc

bottonr of thLr screen. respr:ctively. This inlbnnation retnaincd tln

the screcn unt i l part ic ipants had nradc their choice. We recordcd

part ic ipants ' pr 'ocessing t inre l l 'o tn the pl 'esentat ion o1' the product

i r r l i rnnat ion unt i l they indicated their choice.

Sr.rbsequent ly. par l ic ipal l ts rverc: thankcd for their part ic ipat iot l .

t lchr ic lcd, and () l ' tcrcd t l tc i r contpct ts l t ion.

[irand(lüprcityTHI)Wllran t r,

I latsdu n5 CDs0.003%

I t t

Souy20 CDsO.Olc i .t)

Note . ' l 'HD = to ta l harnron ic d is to r t io l l . Par t i c ipu t t ts in thc h igh eva lu '

abiJity condition rvere also informccl thlt low valucs fitr THD indicatc

bdt tc r so l rnd qur l r ty 1 t1 'p ica l range = 0 .002 ' l to 0 .012" i ) and tha t thc

war ran tv rvas t i len in n ton ths .

tapraid5/zfr 109/zf r2108d09z I xppws I S- l I l1128/08 | 11:55 | Ar t : 2007-0096

COMPA RATIVII TFII n\ K ING

Ij5

Re s tr Lt s u n I L) i.s r: u s.s i ort

Our reasc.rn ing suggests that t l le cf f ic icncy aclvantages ol 'corn-

parat ive th inking are pr i rnar i ly apparcnt i f ' the choice al ternat ives

rure di f f icul t to cvaluate indepenclent ly . because at least sorne of

their at t r ibutcs are k lw in evaluabi l i ty . [n th is s i tuat ion. judges

have to cornpare the choice al tcnrat ivcrs to g ive rneaning to the

othcrrv isc rneani lg lcss inforrnaf ion. ' fh is

compar ison should be

casicr to carry out by participants rvlro lrave bccn proccdurally

pr i rned to engagc in cornparat ivc th inking, so that thc:y should be

faster in rnaking their choicc than contro l part ic ipants. I f thc choicc

al tc ' rnat ives are easy to evaluate indepenclent ly because al l of the

at t r ibutcs arc h igh in evaluabi l i ty , however. no such conrpar ison is

necessary to rnake arr in lbrnrcd choice. Here. pr imhg part ie ipants

on conlpi lat ive th inking docs not hold a processing.rdvantage. In

th is s i tuaf ion. par l ic i l )anls pr inred on cornparat ivc th inking shoul t l

be as last in making their choice as contro i part ic ipants.

As in thc prcvious studi t rs. rve f i rs t excluded response t i rnes that

c i i l icrcd by rrorc than 2. lDs l l r ln the qucst ion nrcans ( l .9c; / r ; o1 'a l l

responscs) f rom fur lher analysis (Bargh & Chartrand. 2000). As a

consc:quencc, latcncics l i r r cr i t ical responscs were not avai lable l i l r; l part ic ipants, such that the analysis of rcsponsc latcncics uas

conductct l on 106 pl r t ic ipants.

Thc rnc:au responss latc:ncics for pal t ic ipants ' choicc, depicted in

Figurc -5. arc consistcnt wi th th is reas<xing. As expected, rvhcn the

provided in l i l rnrat ion was lorv in c:valr . rabi l i ty part ic ipants who

wcrc pr i rncd on conlparat ive processing nrade laster choices be-

twccn the CD changers (M - 10.978 rns. .Sl) - 3.143) than did

contro l part ic ipants (M - 14.423 nrs, SD =. . 4.544), (102) -= 2.50,

I : . 01 , r / . - 0 . 48 . When t he p rov i dcd i n f o rma t i on was h i gh i n

cva lu l b i l i t y . ho r vevc r . t h i s was no t t he case 1M.= 13 ,159 n r s .

.11) =. , .5.9-52 r 's . M -- 12.6f16 ms, 51) . . '5 .4t i0; t < l ) . This pat tcrn

produccd a s igni f icant intc: ract ion c1'1cct in a 2 (conrpar isc ln vs.

coutro l pr inr ingl x 2 (h igh Vs. low cvaluabi l i ty) betwcen-subiects

ANOVA using par l ic ipauts ' choicc latencies as the dependcnt

nreasru-e, I ' ( l . 102) = -1.20, p - .0-1. p l r t i l r l l r = .0; t r . In th is

analysis. nonc o{ thc rcnrain ing cf fccts reached s igni f icance: I <

l , f b r t he ma in e l l ec t o l ' e va luab i l i t y , I ' ( 1 , 102 ) ' - = 2 .40 , p : . 12 ,

lbr the nrain ef fect o l p l inr ing.

I t is notablc that part ic ipants ' cht>iccs betwcen C-D Clranger A

and CD Changer B rvere not s igni f icant ly in l luenced by our

cxpcr inrental r ranipulatrons. ln a 2 (cotnpar isot t vs. conlro l pr in l -

e v a l u a b l e n o n - e v a l u a b l e

F ig r t , 5 . l r r S tudy 7 , mean response la te lc ics lb r thc cho ice l l c twe cn two

br rnds o f CD changers l ry cva luab i l i t y ( l r rgh vs . low) and pr i rn ing (com-

para t ive vs . con t ro l l . I i r ro r -bars represcn l s tandard erLors o f the mcan.

t '7

ing) x 2 (h igh vs. low c:valuabi l i ty) betu,een-subjects ANOVA

using part ic ipants ' choices as a depcndent rneasurc nci ther the

nrain e l lect o1'pr inr ing ( f ' < | ) nor the nrain c: f fect of evaluabi l i ty ,

l ' (1. 106) - 1.74, P : . l ( ) . nor the interact ion ( l ' - ( 1) reached

signi f icance. This f inding is consistent wi th Hsee (1996), who

drr lnonst lated that a s imi lar evaluabi l i ty rnanipulat io l inf luenced

panic ipants ' c l ro ices only i l ' both choicc: opt ions wcrre prcsented

scparatcly, not il they rvcre presentedjointly as was the case in the

present study.

Thesc lesul ts thus establ ish a second boundary condi t ion for the

el l ic iency udvanlages ol ' conrparat ive th inking. Clomparat ive th ink-

ing al lows judges to make choices more rapid ly i l the choice

options are at lcast partly characterized by attributes that are lorv in

evaluabi l i ty .

General Discussion

Thc plcsent rcscarch has exanr incd el t ' ic iency advantagcs o1'

comparat ive in{brn-Lat ion processing. The resul ts of seven stucl ies

clemonstrate the el f ic iency of conrparat ivc th inking. shed l ight on

thc nrcchanisnrs that contr ibute to i t , ancl establ ish boundary con-

ditions for its occurrence. Sttrdies l-4 dculonslrate that compara-

t ive th inking changes in lornrat ion processing. Speci f ical ly , Study

I denronstrates that judges rvho arc procedurally prirned to process

inlbrrnation in a more conrparative lranncr searched for less in-

forrnat ion about the judgrncntal target than did contro l part ic ipants.

Cornparat ive th inking thus induces judgcs to l i rn i t the search lor

target knowledge. Studies 2 and 3 lirrlher show that conrltaretivc

processing induccs juclges to th ink nrore aboul an in lbrrnal ion-r ich

compar ison standarcl about which they act ivate judgrnent-re levant

inlbrnration. Study 4 further suggests that this activated stanclard

knowledge nray bc: uscd as a proxy lirr target inlblrnation that is

r . rnavai lable or d i l f icul t to obtain. These changes in informat ion

processing entai l ef ' t ' ic iency advantages f i r r comparat ive th inking.

In l i le rv i th th is not ion, Studies 2. 4. aud 5 dctnonslrate that

participants who rvcre procedurally prirned kr process informatiort

in a more cornparative manner were more e{'{icient (i.c., fästcr) in

nraking the cr i t ical target judgment. Furthermore, Study 5 shorvs , tq:2

that paf t ic ipauts wlxr re ly more heavi ly on contparat ivc processing

reqnire fewer cogni t ive capaci t ics to solvc the cr i t ical judgment

task. These cf f ic icncy advantagcs of conrparat ivc th inking arc:

part icru lar ly l ikc: ly to occur i l ' (a) the judgrnental target and a

perlinent standard are characterizcd by alignable l'eatures so that a

shared al ignable structure bet$ 'een targcrt and standard is casy t0

cstabl ish (Study 6) and i l ' (b) the leatures ol ' larget and stanclard are

krrv in evaluabi l i ty in that t l rey are di lT icul t to evaluate in isolat ion(Study 7) .

The prcsent research has dentonstrated these elTiciency advan-

tages of conrpalat ive th inking using a var iety of c l i f fcrent mater i -

als. First. wc: used two diffcrrent methods to induce more colnpar-

at ive infornat ion processing. In most studies. we di rcct ly inducc:d

a more or lcss cornparative proccssing style with thc help o1' a

procc:dural pr inr ing task in which panic ipants were expl ic i t ly i t t -

structcrd to eithcrr compare two pictures u'ith one another or to

clescr ibe thcnr separale ly. Stud), 28 used a morc indirect rnethod.

Here. part ic ipunts were pr inrcd on i l more o[ lcss cornparat ive

processing sty le by engaging in onc ol ' two al ternat ive error search

tasks. Part ic ipants had to e i ther repcaledly cornpare a target p icture

to a standard to l ' ind a sct of crrors or they had to c losely inspect

1 8000tt

E roooo. - 14000o.9 rzoooI

g soooo . ^ ^ ^

O 4 n o n

o - ^ ^ ^o z u u ud

0

l 8

tapraid5/zfr-xge/zfr-xge/zlr00109lztr2108d09z I xppws I S:1 | 11/28/08 | 11r55 | Art:2007-0096

]\,IIJSSWI]II'IJR AND t]PSTUDI:

the targct in isolat ion to f ind the s i lnrs erro l 's . Sc:cond. rve used f t rur

di l l 'ercnt typcs of judgrncntal targets. Spcci f ical ly . part ic ipants

. ludged c i thcr speci f ic aspects ofa targct c i ty or a target apartment

or nrade choices betrvee:n different kinds o1' popcorn or dil' lcrent

Cl) changers. Thircl. wc usccl trvo diffcrent ntcasures lbr the

assunred ef ' f ic iency advantagcs ol 'comparat ive th inking. In lnost

studies. wc exaniued how 1'ast part ic ipants rvere i t t making a set o l

judgments. In acld i t ion. rve used a dual- task paradigtn to detnon-

strate that cornparat ivc th inking t rot only a l lows par l ic ipants to

nrakc. iudgnrents l : rs ter but that at lhe santc t ime i t rec;ui rcs less

cogni t ive capaci ty. As in rescarch exarnin ing c l f ic iency advan-

tages in othcf domains (N4acrac et a l . , 1994). th is is appart-nt i r t

improvcd per-formancc in a secondary task. Thc variety ol dilltrent

rncthods wc have appl ied speak to the general izabi l i ty of the

cf f ic iency advantalres ol ' contparat ive th inking.

Limits ot' Elliciertt:t, Atlvuttages

The prcscntcd research has dcnx)llstraled not only that compar-

at ivc th inking has el f ic ic :ncy aclvantages; i t has also c:stabl ished

two bount lary condi t jons. Spcci f ical ly . conrparat ive th inking ap-

pears to a l lorv judgcs to rnakc their judgnrents rnore quickly only

i1 ' targct and stdnclard share al ig i lable featut 'cs and i f they are

cvaluatccl a long di rnensio l ls that arc d i lT icul t to cvaluatc in isola-

t ion. ' l 'h is

ra iscs the qucst ion o1' horv ubic lu i tous the el f ic iency

advantagcs of cornparat ivc th inking real ly arc. In th is lcspcct . i t is

i rnportanl (o note that the cclndi t ions under rvhich cot l rparat ive

thinking provcd to be rnore c lTic icnt arc l ikely to be in p lace under

nrost c i rcLl tnstances. In Studics 6 and 7, part ic ipants wcrcr d i rect ly

providcd with conrparison standards that wc:rc o[ were not char-

acter ized by nonal ignirb le aud evaluable fcatures. Outs ide of the

psychological laboratory. howcvcr. such t l i rec l presentat ion ot

standards that arc characterized by a linlited set oI features that are

expl ic i t ly dcsigncd to be eval t rable ar ld nonal ignablc is unl ikely t<t

exist . F i rst . in many s i tuat ions. . judgcs are not c l i rect ly provided

with a conrpar ison standard but instexd act ivatc th is standard

thcnrsclvcs. Such tncchanisr ls o[ ' s tandard sc: lect iot r are t l f ten

dr iven by structural a l igr tment (e .g. . Musswci ler & Gentner, 2007) '

so that judges are l ikely to uct ivate a standard that shares an

al ignable structure wi th the targct so t l - rat a l ignablc featurcs exist .

LIndcr thesc condi t ic lns. as Study 6 denronstratcs, conlparat ive

thinking holds the dc:scr ibed ef f ic iency advantages. Sccond' targets

ancl standards that arc encounlered ouls ic le o1' the psych<.r logical

laboratcrry i l re typical ly ntore conrplex than the st imul i used in

Studies 6 anti 7 in that they arc not characterized by a tncre four

l 'ealurrrs. but by a larger- o l lent imes secnl ingly indef in i tc : -

nurnber of featurcs. Tl te r r tore featules exist . thc smal lcr t l re

l i k c l i hood t ha t a l l o1 ' t hcm w i l l be eva lL rab l c and nona l i gnab le .

N4()st natural ly t rccurr ing standards wi l l be charactcr ized by a nt ix

ol 'a l ignablc and nonal ignablc: . evaluable and noncval t table {ea-

tures. Studics 6 ancl 7 demonsl tate that i f a subset o1-targct and

stanclard features rvas al ignable and t toucval l table then conlpara-

t ive th inking holds cf f ic iency advantagc:s. What is ntore. wi th

rcspcct to a l ig l lahi l i ty , i1-a l ignablc and nonal ignable f -c l (urcr cx ist .

then . iuc lges typical ly focus on thc al ignable ones (( lcnlncr &

Markman. 199-1: Marknran & Gentner. 199(r) . Together ' th is sug-

gests that the boundary condi t ions we havc: establ ished in Studies(r and 7 are nrorc lhe cxcept ion thrn thc ru le, so that the c l f ic iency

aclvantages ol comparal ive th inking are l ikely to be fa i r ly ubiqui

tous.

A c c u ruct' oJ C r tttt pa rutiv e 7'h irtki rt g

Wc have suggested that of tent intes comparat ive th inking is

norc e f f ic ient wi thout nccessar i ly being less accurate. Clear ly. the

mechanisms of inlortnation ltrcus and informatitln transl'er nray in

principle inl'l ue nce the .luilgmental oulcolnLr. Il orvever. rve contcrnd

that they o{ ien do not . More speci f ical ly , in lbrntat ion locus is

unl ikc ly rc in{ luencc. judgrnental outcontcs i f the i tnpl icat ions of

the in lorrnat ion that . iudges do and dt l not lbcus on dur ing conr-

parat ive th inking are s imi lar . In addi t ion, infbrnrat ion t ranst 'er is

unl ikely to inf - lucnce: judgrnent i l ' the i rnpl icat ions ol ' the t rans-

fcrrcd standard inlonnation fit the target. The latter may tlften be

thc case. because judges typical ly act ivate standards thxl are s im-

i lar t ( ) t l rc target . Consistent rv i th our hypothesis, thc present

f indings do not g ive any indicat ion that cornparxt ive th inking

reduces .judgnrent accuracy. To be sure. wc could not assess

accuracy directly, because the critical judgrnents pertained to fic-

t i t ior . rs {argets 1bl which no t rue valucs existed Di f ferences in

.juclgment accuracy, ltowevc:r, presupposc that the judglnerlts ol'

part ic : ipants who rel ied morc versus less on contparat ive th inking

di t l 'er . ' fh is does not seent to bc the case. as acr()ss the presc: l l t

s ludics no systcnlat ic d i f fcrenccs betwecn the judgrnents made by

both groups arppear to exist . Of coursc. g ivcn the typical atnbigu-

ities associated with the interpretation of null effc:cts. one hils to

use caut ior t in c l rawing l inn concl t ts i t lns l ionr th is evidcnce. Futurc

research rv i l l have to examine how cotnparat ive th inking inf l t r -

ences. judgrnenl accuracy lnore c losely.

Nevertheless, at f i rs t s ight these f indings scem to be at odcls wi th

a substant ia l body ol 'cv idencc that demonstrates that conpar isons

inf luc:uce. iudgnrcntal outcolncs. Rcsearch on judgmental anchor-

ing (Tvc:r'sky & Kahnetnart, 1974), for exanlplc. has denlonstratcd

that nurrer ic est i lnales depc:nd cr i t ical lv on thc compar ison stan-

dards that jutlges are cxposc:d to (for reviews. see Chaplnan &

Johnson. 2002: Musswei ler & Strack. 1999a). Sint i lar ly , socia l

compar isrx research has dcrt ronstrated that sel l -evaluat ions c le-

pencl on tlte social comparistln standards til which judges are

exposcd ( lbr a review. sec Musswci lcr . 2003).

How do these l indings go together wi th the present research?

Wc: suspect that the critical diffcrcnce bctwc:t:tt both scts ol- c:vi-

dence is the rnodc ol s tandard act ivat ion. ln anchor ing and socia l

cornparisor.r research, judgcs arc typically confronted with a judg-

nental s tandald that is c:xpl ic i t ly provided for contpar ison (e.g. . " ls

the pcrcentagc of 'Al ' r ican nat ions in the Uni tcd Nat ions higher or

l<rrver than 650/o '1" "Ärc you morc or less athlet ic than Bi l l Cl in-

ton'1") or is madc accessib le (e.g. . v ia subl in i inal pr imingl Muss-

rvci ler ' & Engl ich. 2005; Musswei ler et a l . , 2004a). Furthetrnorc.

thc:se standards are typically selected to be diffcrent tiotn the

judgrnental targct . In the preseut studies shorving that comparat ive

thinking incrcases judgrnent e lTic iency wi thout lcading to d i t ter-

ences in judgrnent, th is is not the case. Judges are not extemal ly

provided with a standard. but rather, they activate the standards

thcnrselves. Because standard act ivat ion is typical ly led by target-

standard s imi lar i ty (Fcst inger, 1954; Kahnenran & Mi l lcr . 1986) '

juc lges in thc present research are l ikely to use standards that are

sirn i lar to thc targct r i t ther than the diss inr i lar standards that are

typical ly provided. This d i l ' lcrencc in horv s i rn i lar thc used

tapraid5/zfr-xge/ztr-xge/zfr00109lzlr210$d09z I xppws I S=1 | 11128108 | 11:55 | Art:2007-0096

COMPARATIVI] THINKIn.G l 9

compar ison stanclard is to thc judgmental targct lnay be respon-

sib lc lbr the di l ' ferent ia l c l ' fccts on thc iudsnrelr ta l or l tcolne.

Nonntttive and Ef iciencl A.rpccl.r of (.-ontparisrtrt

'fhc plesent pr:rspcrctive on comparison proccsscs supplernents

rnore c lassic perspect ives that have pr imar i ly focuscd on no[n]a-

t ivc aspects o l ' comparat ivc th inking. ' fh is

uonnat ive l i rcus is

part icular ly apparent in t radi t ional v iews o1' socia l cotnl tar isot t

processes (Fcst inger. 1954). Five dccades of rcseat 'ch havc c lem-

onstrated horv strc lngly people re ly on conrpur isons rv i lh others,uvhcn evaluat ing thernselves. Horv people see thenrselves thus

cr i t ical ly depends on how they conrpare wi th othcrs. Fol lowing

Fcst inger 's (195.{) or ig inal proposal . socia l compar ison rescarch

has t radi t ional ly l i rcuscd on thc nonnat ive foundat ions o1'socia l

cornpar ison proccsscs. In part ic t r lar , the funt lamental ussunrpt ion

has long bccn that pc:ople engagc in socia l cotnpar isons to obtain

thc most d iagnost ic int i r r rnat ion lor sel f -cvaluat ion (see Buunk &

Cibbons, 20()7, l i r r a reccnt rcv icw). Supplcnrcnt ing th is t radi t ional

v iew. utr re leccnl evidencc suggests that colnpar isons do not

exclr" rs ively f o l low th is d iagnost ic i ty cr i ter ion. In fact . as we have

pcl int r 'd out bclbre, judges r ' \ 'cn cnsäge in compar isons that are

unl ikc lv to providc thenr w' i th d iagnost ic in lbrrnat ion about their

orvn per lbrrnanccs and character is t ics. Cornl ar ing onc's own per-

fornrancc wi th tha( of a standard who has prcviously received extra

t r a i n i ng i s un l i k c l y t o hc l p cva lua te one ' s own ab i l i t i c s . S t i l l ,

pcoplc spontancously comparc rv i t l t such nondiagnost ic standards(Gi lbcr t et a l . , 199-5). Sirn i lar ly , thc ath let ic abi l i t ies o l 'a profes-

s ional ath lcte such as Michael Jordan hold l i t t lc valuable infornra-

t ion about onc s own abi l i t ies. St i l l , people spontaneously use th is

standard l i r r conrpar ison (Mussrvci ler . Rüter , & Epstude, 2004b),

cven i f 'hc was prcscnted ouls idc o1' their awareness (Musswei ler ct

a l . . 2()04a). Whcreas such f indings arc d i f f icul t to rLrconci le wi th

the nolron that conlpar isons arc carr icd ouf to obtaiu d iagnost ic

in l i rnnat ion about the targct . thcy l re col ls is tent wi th a perrpcct ivc

thzr t lbcuses orr thc c lTic icncy ac. lvantages ol cr lntparat ive procc\s-

ing. Judges nray engagc in colnpar isons wi th i r re levant standarcls

bccausc doing su alkxvs thcnr to t t iake judgmcnts in a quick and

cl l ' ic icnt rnanner. O1' tent imes, cornpar isons nray thus not be pr i -

nrar i ly engaged to obtain d iagnost ic target k l rowledgc but to savc

cogni t ive rcsourccrs in the tatget-evaluat ion process.

f'uturc Direction:;

' I 'hc prcsent conceptual izat ion has a r turrber of intp l icat ions that

nral 'be f ru i t fu l ly cxplored in luture research. For <lne. a structural

a l ignnrent perspect ive suggests fur ther bounclary condi t ions that

rnay boost or rn i t igatc the cf f ic iency advanttgcs o1 ct . rnrparat ivc

th inking. Fi rst . i t has bcen suggcstcd that . judges' ahi l i t -v to cstab-

l ish an al ignable structure depcnds on how tnuch background

knowledgc thcy havc in thcr targct donrain (Centner & Rattcrrmann.

l99l) . For exarrplc. cxpcrr ts arc rnore l ikely ' than novices t<r

t ransfer solut ions betwc:cn structural ly s i rn i lar math problcms

(Novick. l98t t ) . I f comparat ive th inking unfolds along al ignablc

stntcturcs. then th is suggests that thc rnorc background knowlcclgc

antl expertise judges havc. thc nrore they may profit fi 'onr thc:

el1- ic icncy advantages of 'comparat ive th inki r tg. In läct , the elTi -

c icncy of 'cornparat ive th inking may bc a dr iv ing lbrce bel t inc l the

ef f ic icncy of cxpcrt . iudgnrcnt . Second. i t has been dcntonstra led

that t i rne pressure l inr i ts . judges' abi l i ty to cstabl ish al ignable

structures (c.g. . Coldstone & Medin, 1994). To the cxtent that

establ ishing such an al ignable structure is a precondi t ion for com-

parat ivc th inking, th is suggests that judges wi th l imi ted proccssing

rcsources rnay bc less l ikel l ' [o s1u1og in cotnpar ison. The ef f i -

c ieucy advantages of conrparat ive th inking rnay be lcrss l ikc ly to

becone apparent when thcy are most needed.

Furthc:rmore, luturc: research nray supplement the present stud-

ic:s by scrut in iz ing purely proccdural ef f ic iency advantages of

conlpalativc thinking. We have proposecl that the nrechanistns of

inlbrmation locus and information transf'er contlibuto to the ob-

tainccl cfficiency advantagcs. Both o1'thc:se rnechanisnts are inlbr-

mation basc:d. in that they fcrcus ot.t what inlbnnation about target

ancl standard is activatcd and how this infbrnratiort may be usccl to

rnakc comparative thinking nrorc: clTicient. 'fhe

presct.tt research

demoustrates that these inlbrmation-based lnechanislns are ittdcrc:d

at p lay dur ing comparat ive th inking. Morc spcci l ' ica l l ) ' , Studies 1A

aud lB show that cotnparativc thinking leacls.ir.rdges to scrarch for

less in{ornrat ion about the juclgnrental target . Furthcrnrore, in

combination Studics 2A ancl 3 demonstrate that corliparative think-

ing involves thc spontancous act ivat ion of informat ion ubout a

pertincnt stanclard. Study 4 (urther shorvs that this activated infor-

matiorl is transl'erred to thc .iudgntental target. 'fhese

mechanisms

of inli.lrrnation focus and inf<lrrration tt'ansfc:r are Iikely to con-

tr ibutc: to the el l ic iency advantages o1'contparat ive th inking. In

addi t ion to these informat ion-based mechanisms. in lonnat ion-f ree

mechanisms thirt opc:r'ate without the activxtion and usc: of specif icr

inlbrmation ahout the target and the standard rnay also be at play

dur ing conrparut ive th inking. Future research rv i l l havc to ident i f ,v

whether infornration-liee rnechanisrtts contribute to the efficiency'

advantagcs of conrpat 'at ive th inking and how they interact wi th thc

information-based ntechanisnls that werc identiäed in the prc:sent

rr:sealch,

In addi t ion, our reasoning ra ises the quest ion ot-how cotnpara-

t ive th inking inf luences sat is l iLct ion. We have suggested that com-

parat ive th inking al lows . iudges 1() makc . iudgnrents and choices

mure quickly. of tent i rnes wi thout obtain ing worse outconres. But

are ir"rclgcs rvho rcly n)ore on comparative thinking also as happy

with thc judgrnent or choice thcy ltave nrirde as .ittdgc:s who rely

less on corrrparison? So lar. we have ntlt examined horv cottrpar-

at ivc th inking inf luences sat is tact ion. Research on choice over load

(e.g. . Iyengar & L.eppcr, 2000; Schwartz et a l . . 2002), horvever,

suggests that part icular ly in s i tuat ions in which people are over-

wl ic l rned by their choice opt ions, comparat ive th inking nray le-

duce judges' satisläction rvith tl-reir ou(come . Too nrany options in

a choice s i tuat ion lead to decreased sat is fact ion wi t l l t l te actual

outcome ( lyengar & Lepper, 20001 Schwartz et a l . . 2002). This is

part icular ly the case 1bl judges who t ry to lnake the best possib le

choicc (Schwartz et a l . , 2002). I l is notablc that th is lat ter tcndency

is re lated to judgcs' rc l iance on socia l contpar isons, which sug-

gests lhat judgcs who 1ry to make t l te best possib lc choice rc ly

rnore hcavi ly on comparat ive th inking i t t gc: t teral . In th is rcspect .

these I ' indings are a f i rs t . adrni t tcdly indirect , h int that contpalat ivc

th inking rnay at t imes lead judgcs to be less sat is f ied u ' i th thci r

choice. Futurc research wi l l have to examine how c()mparal ive

thinking inf lLrences sat is lact ion more di rect ly .

I t is a lso intcrcst ing thal thc cogni t ive ancl af fcct ivc conse-

qucnces of not being sat is l ied-counter{actual thoughts and the

expc:r i r :ncc: o l ' regret-are also c losely l inked to compirr i ron

20

lapraid5lzlr-xgelzfr-xge/zfr00109lztr2101d09z I xppws I S=1 | 11l2Bl0B I i1:55 | Art: 2007-0096

M USSWIiII-llR AND IIPSTUDIT

processes. namcl1, , bclween the actual and an al te lnat ive out-come (scc: E1;stude & Rocsc. 2008. and Zeclenberg & Pieters,2007. l i r r reccnt l 'ev iews.) . The elTccts of contpnln l lye in lbrnra-l iun proccssing can t l rercforc: be expected to go far beyond thcel ' f ic iency of inforrnat ion processing, . judgnrent . and choice.I ns tead . i t s c :ems na tu ra l l o assu lne t ha t j udgcs ' c va lua t i ons o1 't hc i r j udgn rcn t , dec i s i on . and cho i cc ou t con l cs w i l l a l so beshapcd by con rpa ra l i v c t h i nk i ng .

Coucl t t .s ior t

Wc spcculatc ' t l la t l l lc c l f ic iency aclvantages of cornpurr t ivcth inking may be one reason l i r r thc apparent ubiqui ty of compar-isons in hr , rman inf i r r r lat ion processing. No r lat ter whether pcoplepcrccive, pfocess. and cvaluate int i l rntat ion ahout s i rnplc: phvsicalobjects or crrmplex socia l s t i r ru l i . thcy rc ly on cornpar isons wi th apcrt inent nonn or standard (Kahncrnan & Mi l lcr , l9t l6) . In fact .th is leaning toward cornparat ivc th inking is so strong thal judgcs

cvcn rnake use o1- i r rc: levant and nr is leading standarcls. In l ight o1'thc prescnt l ' indings. conrpar isons ntay play such a pivola l ro lebecause thcy al low judgcs to obtain an ef l ' ic iet rcv-cuhancing in-fornat ional l t lcus in pmcessir . rg target informat ion. Cumpar isonsarc thus-abovc: a l l - rc lat ivc ly last .

References

Ahelson, f t . P. (1995). Sld l l r / l ( r . r u. t l t r inL. i l t lu l urgumei l t . I l i l lsdale. NJ:[ : r lbauni .

Ardersel . S. M.. & Chcn.,S. (1002). ' l 'he rc lat ional sc l f : An interpcrsorralsocir l -cogni t ivc theory. I ' . r r r ' / ro lagicr i l /?er lc ' r r . /09, 619.-645.

Bai l largeon. I l . ( l99l ) . l lcrsoning about the hcight and locat ion ofa hiddenobjcct in -1.5- and (r .5-mol th-ol t l in fants. Cogni t ion. J8, l3-42.

ßa l l ] h . J . A . ( 1981 ) . A l t en t i on r r r d au to rna t i c i t v i n t he p roccss ing o fsclt--rcievrrnt inlbrmltion. Jounutl o.f Par.ronuLitt ttttd So<'iul /'slclla/r;gr,.1 . 1 . . +25 436 .

i largh. J. A. . & Chartrand. T. t - . (2000t. The rn ind in the middlc: Apract ical r r r ide to pr i r r r ing and autourat ic i ty research. In I -1. T. Reis & C.Judd tt:ds.), Ilandltook o.f rcsaax'h ntt'thol.s itr social arul personttlitt'

Tr . r lcha/o,gr ' (pp. 25.1-285). Carnbr idge. MA: Cambridge Univers i tyPrcss.

Iliernat. l\1. (2003). 'lixvarrl a broatler vicrv o{ social stereotypinr. Arr(,r-

icun Pst< l to logisr . .58, l0 l9-1027.I l icrnat . N1. . & f : rdclman, S. (2007). Standards. In A. W. Kruglanski &

l:r. T. Iliggins (Llds.). "Srrllal /rs-rcho/og: Hundbook of busit principle.st2nd ed. . pp. 308-333). Nes,York: Gui l lbrd Press.

l lodcnhluscn. G. V. (1990). Strreotypes as judgmental heur ist ics: [ ]v i -t lence ol c i rcadian var iüt ions in d iscr iminat ion. I 's lc lut logicul St ience,/ . - r l 9 32 : .

Brrc lcnirauscn, Ci . V. , & Lichtenstein. M. ( 1987). Socia l s lereotypes andinfonrrnt iorr-proccssing strategies: T 'he impact of task corrrp lexiß. , lour-ttttl f Par.xntulit.t rrrul Social Psvhologt, -52, tiTl--880.

Brown. D. R. ( | 953). St i rnulus-s i rn i lar i ty an( l the anchor ing of subjcct ivescales. Ärrcricurr Jourtnl oJ Pst'tltolot.t. (t6, 199-:l;1.

Bruner. J. (1957t. On perceptual readiness. P.sytholoSi ta l Review. 6.4,I l3- | 52.

Buunk. A. P. . & Gibhons. I r . X. (2007). Socia l conrpar ison: The end of rtheorl rnd the ernerge ncc oi a ficld. Or.quni:.utiottrtl llelnt'irtr andIttrrttrttr [)eci.sirnt ['rocesscs. 102. 3 2l .

Chapnran. G. I l . . & Johnson, 11. J. (2002). l l lcorporat ine the i r re lcvant :Anchors in judlncnts o l bel ie l and value. In- l ' . Gi lov ich. D. Cjr i f f in . &D. Kahnctnan (Eds.). //('xri.r/i( s anrl bittst.s:

-l-he ltrt,chologt of intuitive

. ju lqntut t (pp. 120- l l l i ) . C'arnbr ic lcc. N' [A: Clantbr i r lge Lln ivers i ty Press.

Chopl in, J. M.. & Hurnnrel . J . U. (2002). Magni tude compar isons distor lnren(al representations of rrragnitutlc. Journul ol Expet'inrctüul Pstthol-ogt: Cerrcral, l.l1, 270-286.

Chopl in. J. 1\4. . & Hunrrrre l , J . 11. (2005), Compar ison- induced decovelfecls. lulentot .t' & Co;4nition, -?.1, -l-12-341.

Clement. C- A. . & Centner. D. (1991). Systctnät ic i ty as a select ion col-st ra int in analogical mapping. ( 'ogt t i t ive.Sclerce, 1-5, 89-132.

Coren. S. , & Errrrs, J . T. (1993). Sizc contrast ls a tunct ion of f igurals imi lär i ty . Parc 'c1t t ion und Psvrhol t l t ts ics, 54, 579 588.

I )ehaenc' . S. . Naccache. [ - . . [ -c Clec 'H, Ci . . Koechl in. I ] . . Muel ler . M..Dehaene- l -arnbcrtz, Ci . , et a l . ( lc)98. Octnber 8) . l r r ragine urrconscioussernantic priming. Nuturt, -195, 597-600.

Dunning, D. , & Hayes, A. I r . ( 1996). I l , idence of egocentr ic compar ison insocial jurlgment. .lounrul oJ I'ersonality nnd SociaL I'st'chologt, 71,213-229.

I :ngl ich, 13. , Musswei ler . T. , & Strack, F. (2006). Playing dice rv i thcriminirl sentcnces: The influence of irrelevant anchttrs rtn etperts'judicial decisiorr nraking. Pcrsarclity rtttd Soc ul Pstchologt Bulletin,.12, I 88-200.

l lpstudc. K. . & Roese, r .* . J . (2008). The funct ional rheory of counter factualth inking. I 'er :onal i t . t and Sot ' io l Psy<:holot t Reyi t tv . I2, I68,192.

I:estinger, L. (19-5-t). A (heory of social comparison processes. llrrncrrRelutiuts, 7, lll-140.

Gelman, S. A. , & Mrrkrnan, I i . N{, (1987). Young chi ldren's induct ionslrotn natural kinds: l'he role of categories and appeararrces. C/rllr/Davclopntent , - t8, 1532.-1541. AQ:3

Ger) tner, D. . & Markman. A. I l . (199.1) . Structural a l igrrrnent in cornpui-son: No di tTelence wi thout s inr i iar i ty . I 's tchological . i ' ierrce, J, 152-t - s8 .

Gentner, D. . & Marknran, A. B. (1997). Structure mappirrg in analory anclsi mi lari ty. At n t ri t a n P.s.t' t l t olo t i s r, .52, 45-56.

Gentncr, D. . & Medina. J. (1998). Simi lar i ty and the development of ru les.Cognit ion. 65. 1.63-297.

Cent l lcr , D. , & Ratternann. M. J. ( l9r) l ) . l -anguage and the career ofs i rn i lar i ty . I r r S. A. Gelman & J. P. I lyrnes ( [ ]ds.) , Pzrr l r r t t i r r . . r , r rlllonshl und lunsu.tg(: Interralntions in tlevelopntent (pp. 22-5-277).l -ondon: Cambri r lge Univers i ty Press.

Ci ick. M. L. . & Holyoak, K. J. (19110). Analogical problern solv ing.Cogttitive Pst,chologt'. /2, 306-355.

Ci i lbert . D. T. , Gieslcr . I t .8. , & Morr is . K. A. (1995). Whcn comprr isorrsarise. JurnttrI ol Personu|i|v ttrtd Sociul Pst,t:Itoktgt, 69,22'l-236.

Cjol t ls tonc, R. L. , & Mcdin, D. L. (199.1) . T ime course of compar ison..lournal ol E4terinrcntul Ps,tcÄo/oo. ltarnint. ülc'uutn'. und Cogni-t ion, 20,29-50.

['le lson, ]1. (1961). Aldptdtirn level thtort': An ttpe rin<'tttol and .rt.ttt,trr-trtic trltprourlt to helnvior. Nerv York: Hrrper.

[ ' lcr r . [ ' . N{. (198(r) . Consequcnces ol pnming: Judgment and l rehavior .JotrrnaI o l Per. tonül i t \ , . tn( l Socia l Psrr :holog,r ' , 51, I I06- l I |5.

Fl iggins, I1. T. (1987). Sel l ' -d iscrepancy: A theory re lat ing sel l and af l lc t .P s.rr lu;lositu I Rn,ior', 9J. 3 I 9-3.+0.

Higgins. tJ . ' l ' . , & I -ur ic , L. (1983). Cortext . categor izat iur . ancl recal l : The"changc-ot'-standard" cfTect. C o gni I i t t, P st c h o I o {ay, I 5, 525-5 47 .

Higgins. E. T ' . , c t ,Stalgor, C. (1988). A "change-of-standard' pcrspect ivcon fhe relations illong contcxt, judgrnent, and rnernory. .lournal of'PersrnoIity and Socitrl Pslc/iolo,g.r', -5.1. |8I-I92.

Higgins, E. T. , Strauman. ' I . .

& Klein, I f . ( 1986). Standards and the processol 'sel f -evaluat ion: Mul t ip lc af tects f |om mult ip le stages. In R. M.Sorrentino & li. T. Higgins ll1lJs.). Iluntlbrxtk of nntivttion und togni-tion: Foundatitnrs oJ so<ial bthavior 1pp. 23-63). Chichester. UnitcclK ingdom: W i l ey .

Hsee. C. K. ( 1996). Thc evaluabi l i ty hypothcsis: An explanat ion of prel ' -erence tevers l ls bctween jo int and separate evaluat ions of a l tcrnat ives.Orguni:ationul llehot'ior and llumnn Dcr'ision Processes, 67, 21'l-25'7.

tapraid5/zfr-xge/zfr-xge/2fr00109/zfr21OBd09z I xppws I S=-1 I 11l2Bl0Bl 11:55 | Art:2007-0096

COMPARATIVI l Tt l l N KI t \ . .G 2 l

I vengar . S . S . . & t -epper . N ' [ . (20(X ' ) ) . When cho ice is dcrno t iva t ing . Joarna l

oI PcrsontrIi |t untl Sot:iuI Pst't:|toIogt'. 79, 995--|l\t\6.

Kahneman. D. . & Mi l le r . D . T . ( 198( r ) . Norm theory : Cornpar ing rea l i t y to

i ts a l tc rna t ives . P . t tchr t log icu l Rev ievv , 9 -J , l3 ( r -153.

Krucger ' , J . (1998) . On the pcrccp t ion o l soc ia l conscrsus . Advunces in

E.r pt ri nt e r üu I So c i tt I P.s 1' c ho lo gv. .J0, I 6.1 -240.

Kruegcr. J. L (2007). Irrorn social projection to social br.havirn'. Europcurt

Ret , ie t o l So t io l P , lc l to /o t r , , /8 , l -35 .

Krug lansk i . ; \ . w . ( l9 fJ9) . The psycho logy o f be ing " r igh t " : The prob lenr

of accurrcv in social pcrc:eption ancl coqnition. Pst'clutlogical Bulletin,

/ ( /ö ,395- "109.

\ ' [ac rae . C. N. . & I ]odenhauscn. G. V . (20 tX) . t . So t ia l cogn i t ion : Th ink ing

AQ: .1 c r te to r ica l l y about o thers . Annuu l Rt ' r ' i c ty o l Ps t '< :ho logr , .5 / . 93-120.

N ' lac rae . C . N. . l r ' l i l ne . A . 8 . . & l lodenhauscn, G. V . ( 199.+) . S te reo types as

er rc rgy-sar , ing c lev ices : i \ peek ins ide thc cogn i t i ve Loo lbox . J t ru rn t l o l

[' t r's o t n l i t v u t d S o t i u l l'.rlc /to1rr.gr', 66, 37 -'+]

N larkman. A . D. . & Gcntner . D . (1993) . S t ruc tura l a l ignment dur ing

sirnilarity cornparisons. Cognitivc / 's.lr ' /ra1oo', 2J, .11 l-467.

Markrn ln . A . D. . & Gentncr . D . (1996) . Cornrnona l i t i cs and d i f f c rences in

similarity corrrparisons. Mtnton & Cognitiort, 24, 235-249.

N lcd in . I ) . L . . Go lc ls tonc . I l . . L . . & Gent t rc r . D . (1993) . Respects fo r

sirnilarity. P.rvtlntlopicul l levievr, I 00, ?54*276.

Mi l le r . I ) . T . . & Prent ice , l ) . A . (199( r ) . T l rc cons t tuc( ion o f soc ia l norms

lrnd stanthltls. ln I i. ' f.

Higgins & A. W. Kruglanski (Eds.), .Socizr/

ltstcholoq.r: l l ttttdlnrtk ol basic princilt les (pp. 799-ti29). tr*erv York:

Gu i l f i r rd Prcss .

N ' [o rse . S . . & C iergcn. K . .1 . (1970) . Soc i l ) compar is r ) r ] . se l l -cons is tcncy .

artd the conccpt of sclf. ./rrrrrrrri l tt l Parsotnli l t ' and Stxiul f 'sttholrtgt',

/ 6 . l + 8 - - t 5 6

Mussrvc i l c r . T . (2001) . "Seck and yc sha l l l ' i nc l " : An teccdents o f ass imi -

fa l ion anc l con t ras t in soc ia l compar iso l . I iu ro l teun. lourna l r t l S tn ; iu l

I,,rr..holacr,. -l I , 199 -.509.

Mussrvc i l c r . T . (2001) . Co lnpn ison processes in soc ia l judgment : Mech-

lunisms antl conscquer)ccs. P st c ltologictt I l let ' icv. I I 0, 4'72- 189.

N ' lusswe i l c r . T . . & I lodcnhauscn. G. V . (2002) , I know you are bu t what anr

I? Se l f ' -eva lua t ive cünsequences o f judg ing i t tg roup and o t l t s loup l t l cm-

,\Q:5 hers. ./orrrnal o.[ Per.sonalitt cttul So<'iul I 'sv'holoct', 32, 19 -32.

N ' [ussrve i le r . T . . & [ :ng l i ch . I ] . (2005) . Su l r l im ina l anchor ing : Judgrnenta l

consequcnces an t l u r rdcr ly ing t t tech ln is r t rs . Or l i tn i :u l iono l Behut ' io r

urul I lumtut [)tci.siort Proces.ses, 9,9, | 33..- 1.13.

N lusswei lc r . T . . & Centncr . D . (1007) . On app les and ora t tges : S t r t t c tu r i t l

a l ignr r rcn t in thc sc lcc t io r r o f soc ia l c ( )mpar ison s tandarc ls . Joru t tu l o f

Cogr t i t i r ,e .Sr ' i c r r r :e . 8 , l -38 .

N lussweihr . T . . & l l .ü te r , K . (2003) . Wl ra t f l i cnds are fb r l Thc use o l '

rou t i r re s tan thrds in soc ia l compar ison . . louu t t : t l t t . f I ' e r .son t t l i l t ' t t r t r l

Sociul I 'stchrtlogr', 8.5, "167-'18 L

lV{ussrve i l c r . T . , l l ü tc r . K . . & Eps tude, K . (200.1a) . ' I ' hc rna t t who rvns t t ' t

there : Sub l im ina l soc ia l conrpar ison s tanc la r r l s in f luence sc l l ' - cva lu l l . iu r ,

Journul ol E.rpt't ' intental Sociul Psvcholott, 40, 689-696.

lv lussq 'e i le r . '1 . . R t i te r . K . , & l ips tude, K . (200 '1b) . The ups anc l dorvns o l

soc i r l cor t tpar ison : l v lechan is tns o f ass imi la t ion Jnd cont ras t . Jo t tn ta l t t l

PtrsrtnLtIitt uttcl SocIttI P.st'cluslo.qt, 82, 832-8.14.

N ' lussu 'c i le r . ' l ' . . & S t rack . I r . (1999a) . Cornpar ing is be l i cv ing : A se lcc t i ve

rcccss ih i l i t l " mr r t l c l o f . iud .urnenta l anchor i t tg . ln W. S t roebe & lv l .

Hcws(one (Eds.J. Europeurt rt 'r icu' ol .srtt itt l pstrhologt' 1Vol. l0' pp.

135 167) . Ch ichcs tc r . [ -1 : r i ted K ing t lonr : Wi lcy .

l v l r rss rve i le r . T . . & St rack . t r . (1999b) . I l ypo thes is -cons is ten t tcs t ing and

scmünt ic p r i rn ine in the ancht t r ing parad igm: A sc lcc t i ve access ib ih ty

tnoc le l . Joar r ta l o l E . r l te r in tn tu l 9 t t iu l Ps t tho l t tg r ' , -15 , |36- |64 .

Navon. I ) . . & Gophcr. D. ( 1979). On the econonry of thc hurnan proccssing

systern. Psr,r:/ra1o g i c u I ll t t' i c tr'. I 6, 2 | 4 -25 5.Neely. J. H. 11977\. Seruantic prirrring arrd rctricval from lexical nrcmrrry:

Rolcs of inhib i t ionless spreading act ivat ion and l imi tet l -capaei ty r t tet t -tion. .l o u r n tt I of E.r p c r i m e n l a I P s y c ho k:; g t', I 06, 226 -25 4.

Nosot.sky, R. M. (1986). At tent ion, s imi lar i ty , ant l the ident i l lcat ion-cxtegorization relationship. Jounnl ol E.rperintental Pstchologl': Gen-e r a l . 1 1 5 . 3 9 - 5 ' 1 .

Nosot,sky, R. M., & Palmer i , T. J. ( 1997). An exemplar-based randtxt t walk

rrrodel of speedcd classilication. P.rlcÄologitol Rct,it'tv, /0J. 266-300.Novick. L. l t , ( l9 l t8) . Analogical t ransfer , problcm simi lar i ty . and exper-

tise. Journul ol ll.rperinrcntt:rl Pr'1,c7t,r1rt.,.' Leurnittg, Mcnnrt', und Cog-n i l bn . l 4 . . s l 0 -520 .

Ross. L. . Grecn. D. , & Housc. P. ( | 977). The fa lse consensus phenomenon:

An attributional bias in selt-perception rnd social-perception processes.

Journal ot' Etpe t'imcntol Socinl P.sttlrclog,t, /.i, 279-301.Ri i ter , K. . & Musswei ler , T. (2005). Bonds ol ' f r iendship: Cotnparat ive

sel f -cvaluat i r r r rs evoke l .he use o{ rorr t ine standarcls. Sot ia l Coqt i t ion,23.l 37- r60.

Schwartz, I l . , Ward, A. . Montcrosso, J. , l -yubomirsky. S. , Whi te, K. . &Lehman, D. R. (2002). Maxirn iz ing versus sat is f ic ing: Happincss is r

rnattel of choicc. Jolrntcrl ol Persorutlilv ttntl Social Pslftokrgt', 8-1.

I t79 -t 197 .Shcr i i , M., & Hovland. C. t . (1961). Sot i t t l ju lgntcnt : Assinr i l t t i t tn t tnd

conlr(1st effects itt communicution und ultitude cltartgr'. Ncrv Havcn. CT:Yale University Prcss.

Sherman, J- W., Kle in. S. 8. . l ,askey, A. . & Wyer. N. A. ( 1998). Intergroup

bias in gror tp judgment processes: The ro le of bchavioral t t tentor ies.

Jotntul ol E.xpet'intenlttl Sotiul I'stchologt, -14, 5l-6-5.Sheman, S. J. , Ahlrn, D. , l3crrnan, t - . . & Lynn, S. ( l97f i ) . Cot t t rast cf fccts

and their refationship to subsequent bchavior. .lournol oJ Erlrerinlcntdl

Sotill I'svholoc,r, 14, 340-350.Sherman. S. J. , Houston. D. A, , & I iddy. D. (1999). Canccl lat ion and focus:

A fcature-uratclting rrrodel of choicc. In W. Stroebe & IVl. Hervsttlne( l )ds. \ , Europeat l vs l t isÄ, ( t socie l pst<: l tohtgl (Vol . 10, pp. 169-197).

Chichester , Uni ted Kingdorn: Wi ley.

Snr i th, U. R. (199.1) . Procedural knowlcdge and processing s l r i r lcs ics in

socia l cogni t ion. ln R- S. Wyer & T. K. Srul l (ht ls . ) . I landhook r t f sot : iLr l

utgni t i rnt (2nd ed. . Vol . l . pp. 99-1,52). I ' l i l lsdalc, NJ: Er lbaurn.

Smith, I l . R. , & 7-aralc, N4. A. (1992). Exernplar-bascd nodel of socia ljutfgment. Pst<'lutlogicul /ler,ft'rr. 99, 3--21.

Swrrrr r . W. B. ( 1984). Que st for lccuracy i t t person perccpt ion: A mattcr ofpragnatics. P.rt'cholociutl Ret'it'tv, 9 l, 15'7-111.

' l 'ay lor , S. 11. ( lg l t l ) . T 'he inter face ofcoqni t ive and soc: ia l psy-chology. In

J. Harvey (.F.tl.). Cogttitiort, sotial hehat,ior', and tlte envirotuntnl (pp.

182.-21 l ) . Hi l lsdale, NJ: I i r lbaum.Tversky. A. , & Kahneman, D. ( lc)7.1, Septenlber 27). Judgment under

uncerta inty: l leur is t ics and t r iases. Scicncr, /8.5, I l2 . t - l l l0 .

Wickcns. C. D. (1976). The ef fects of d iv ided at tent ion in in lbrrnat ionproccssirtg in trtcking. ,lournttl of E.rperint<'tttttl I's1'thrtlttgt: Iltunutt

Perctption anrl Pt' /itrntontc. 2, l -l '3.

Zcelcnberg, M., & Pictcrs. R. (2007). A theory of rcgrct rcgulation. .ltturnal

of Consurner Psrclnlöev, 17, 3-18.7-halg. S. . & Marknran, A. I l . (2001). Proccssing product uniquc lcatures:

Af ignahi l i ty and involvet t tcnt in prefcrence ct lnstruct ion. Journal r t l

Consutner P.stc:holotv, I 1, l3-71 .

Rece ived Ju ly 3 ,2007Rcvision receivcd September 18. 2008

Accepted Scpterrber 22. 2008 I