Table Of Content

JudgmentandDecisionMaking,Vol.5,No.4,July2010,pp.285–299 Physiological arousal in processing recognition information: Ignoring or integrating cognitive cues? GuyHochman∗ ShaharAyal Technion—IsraelInstituteofTechnology DukeUniversity AndreasGlöckner MaxPlanckInstituteforResearchonCollectiveGoods Abstract The recognition heuristic (RH; Goldstein & Gigerenzer, 2002) suggests that, when applicable, probabilistic infer- encesarebasedonanoncompensatoryexaminationofwhetheranobjectisrecognizedornot. Theoverallfindingson theprocessesthatunderliethisfastandfrugalheuristicaresomewhatmixed,andmanystudieshaveexpressedtheneed forconsideringamorecompensatoryintegrationofrecognitioninformation. Regardlessofthemechanisminvolved,it isclearthatrecognitionhasastronginfluenceonchoices,andthisfindingmightbeexplainedbythefactthatrecog- nitioncuesarouseaffectandthusreceivemoreattentionthancognitivecues. Totestthisassumption,weinvestigated whetherrecognitionresultsinadirectaffectivesignalbymeasuringphysiologicalarousal(i.e.,peripheralarterialtone) in the established city-size task. We found that recognition of cities does not directly result in increased physiolog- ical arousal. Moreover, the results show that physiological arousal increased with increasing inconsistency between recognitioninformationandadditionalcueinformation. Thesefindingssupportpredictionsderivedbyacompensatory ParallelConstraintSatisfactionmodelratherthanpredictionsofnoncompensatorymodels. Additionalresultsconcern- ingconfidenceratings, responsetimes, andchoiceproportionsfurtherdemonstratedthatrecognitioninformationand othercognitivecuesareintegratedinacompensatorymanner. Keywords:affect,recognitionheuristic,physiologicalarousal,parallelconstraintsatisfaction,noncompensatorymodels. 1 Introduction typical tools in this toolbox, suggests that probabilistic inferences about a specific criterion (e.g., which of two Imagine a business trip to a city you have never visited citiesismorepopulated),aremadeonthebasisofrecog- before. Your meetings end later than you had expected nitionalone,sothatrecognizedobjects(e.g.,thenameof andyoudecidetospendthenight.Youmakeafewphone thecity)willbechosenoverunrecognizedones. callsandfindtwohotelswithvacancies,bothatasimilar Indeed, it has been shown repeatedly that people ac- rate. Youknownothingelseaboutthesehotels;however, tuallypreferrecognizedobjectsoverunrecognizedones, youdorecognizethenameofoneofthem. Whichhotel and that such behavior may lead to rather accurate in- wouldyouchoose? AccordingtoGoldsteinandGigeren- ferences (e.g., Borges, Goldstein, Ortmann, & Gigeren- zer (1999; 2002), these kinds of probabilistic inferences zer, 1999; Goldstein & Gigerenzer, 1999, 2002; Pachur couldberesolvedbyrecognitioninformationalone. & Biele, 2007; Pohl, 2006; Reimer & Katsikopoulos, More specifically, according to Gigerenzer and col- 2004;Serwe&Frings,2006). Nevertheless,severalstud- leagues (Gigerenzer, 2000; Gigerenzer & Goldstein, ies challenge the claim that recognition information is 1996; Gigerenzer, Todd et al., 1999), individuals are used in a noncompensatory manner1 as predicted by the equipped with several tools (i.e., fast and frugal heuris- RH(e.g., Bröder&Eichler, 2006; Hilbig&Pohl, 2008; tics) that exploit the structure of the environment, some McCloy & Beaman, 2004; Newell & Fernandez, 2006; of which rely solely on one piece of information at a timeontheroadtoreachingadecision. Therecognition 1Noncompensatorystrategiessuggestthatdifferentcuesorattributes ofthechoicealternativeareconsideredinaspecificorder,usuallyin heuristic (RH), the simplest and one of the most proto- theorderoftheirvalidityorimportance(e.g.,thetakethebestheuristic, TTB).Onlyonepieceofinformationisbeingevaluatedatatimeand ∗Address:GuyHochman,MaxWertheimerMinervaCenterforCog- cuescannotcompensatefordeficitsinotherscues. Incontrast,com- nitive Studies, Faculty of Industrial Engineering and Management, pensatorystrategiesassumethatchoicesarebasedontheintegrationof TheTechnion-IsraelInstituteofTechnology, Haifa, Israel. E-mail: theavailableinformation.Thus,compensationbetweencuesispossible [email protected]. (e.g.,theExpectedUtilityTheory—EUT). 285 JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 286 Newell & Shanks, 2004; Oppenheimer, 2003; Pachur, —akintoperception(Holyoak&Simon,1999;McClel- Bröder&Marewski,2008; Richter&Späth, 2006). For land&Rumelhart,1981;Read,Vanman,&Miller,1997; example, in Oppenheimer (2003), Bröder and Eichler Thagard,1989)—bywhichconsistentinterpretationsof (2006), and Richter and Späth (2006), additional infor- decision tasks are constructed. Available information is mationhasbeenfoundtoaffecttheproportionofchoices taken into account according to its importance, and the thatfollowedrecognition. Whileithasbeenarguedthat advantages of the (emerging) preferred option are high- several of these studies suffer from methodological lim- lighted by systematic information distortions (see also itations(seePachuretal., 2008), theirfindingsnonethe- Russo, Carlson, Meloy, & Yong, 2008). Theformalized less highlight the need for considering a more compen- PCSmodelincludesfourpossiblesteps. First,whenpre- satoryintegrationofrecognitioninformation. sented with a decision task, related information is acti- Moreover, studies on the RH have focused mainly vatedtoformamentalrepresentationofthetask.Second, on choice behavior. Yet, since the choice predictions automaticprocessesofPCSleadtomaximizationofcon- made by different models often overlap (see Ayal & sistency by automatically highlighting information sup- Hochman, 2009; Glöckner & Betsch, 2008a), it is dif- portingthefavoredalternativeandatthesametimesup- ficult to draw clear conclusions concerning the underly- pressingcontradictinginformation. Asaresult,aconsis- ing cognitive processes. Several recent studies have at- tentrepresentationoftheavailableinformationisformed. tempted to address this problem, for example by testing Third, the decision maker evaluates the choice alterna- hypothesesconcerningprocessmeasuressuchasdecision tives, and if one alternative clearly dominates the other time(Hilbig&Pohl,2008;2009;Marewski,Gaissmaier, (i.e.,ifthereisenoughinformationinfavorofthisalter- Schooler, Goldstein, & Gigerenzer, 2010), and by addi- native), this alternative is chosen. However, if informa- tionallyinvestigatingconfidence(Glöckner&Bröder,in tionconsistencyfailstoreachacertainthreshold,thede- press). Yet, additional research is needed to disentangle cisionmakertakesaforthstep, inwhichdeliberatecon- theroleofrecognitioninformationindecisionmaking,as struction processes are activated to change the structure wellastoexaminethecognitiveprocessesunderlyingits ofthenetwork(e.g.,changingthepatternofinformation useininferencesandchoicebehavior. search,assigningdifferentweightstothedifferentpieces To achieve these goals, we investigated the nature of of information). Among other things, the PCS model is the processes underlying the use of recognition infor- supportedbyresearchthatshowsthatunconsciousmod- mation in probabilistic inferences, using physiological ificationsoftheavailableinformationoccurduringdeci- arousal, choice proportions, decision times, and confi- sion processes (e.g., Glöckner, Betsch, & Schindler, in dence as dependent measures. We derived concurring press; Holyoak&Simon, 1999; Simon, Krawczyk, Ble- predictions on the basis of a noncompensatory perspec- icher, & Holyoak, 2008; Simon, Snow, & Read, 2004). tive, and juxtapose these with contradicting predictions Furthermore, PCS models have been found to account derivedfromacompensatorymodel. Specifically,wein- well for both choices and decision times in probabilis- vestigate whether recognition leads to increased physio- ticinferencetasks(Glöckner,2008;Glöckner&Hodges, logical arousal, which dominates other cognitive infor- inpress;Glöckner&Bröder,inpress;seealsoGlöckner mation, as suggested by affect-based noncompensatory &Betsch,2008c). models(e.g.,Damasio,1994;Slovic,Finucane,Peters,& MacGregor, 2002). It has been argued that recognition 1.1 Overviewofthestudy information,likeotheraffectivesignals,mightbegener- ated automatically (Pachur & Hertwig, 2006), and that Thecurrentstudyisdesignedtoexaminewhetherrecog- itmayoverridecognitiveinformation(e.g., Goldstein& nition information is integrated with other information Gigerenzer,2002,Experiment2;Pachuretal.,2008,Ex- in a compensatory or noncompensatory manner, and periment1). Thus,weexaminedwhetherrecognitionin- whether recognition information is encoded as a special formationisaffectivelyencodedduringthedecisionpro- affective signal. To do so, we examined choice behav- cess. ior, while participants conducted a city-size task (Gold- We tested the RH against the compensatory Parallel stein & Gigerenzer, 2002; Richter & Späth, 2006). The Constraint Satisfaction (PCS) model for decision mak- city-size task was chosen in order to create an optimal ing (e.g., Glöckner & Betsch, 2008b). The PCS model environment for using the RH. Specifically, in this task was selected because it has been shown to account well thecriterion(i.e.,populationsize)isnotimmediatelyac- fordecisiontimes(Hilbig&Pohl,2009)andconfidence cessible, and inference has to be based on probabilistic (Glöckner & Bröder, in press) in recognition tasks, and cues(Goldstein&Gigerenzer,2002;Pachuretal.,2008). because it makes specific predictions concerning physi- Inaddition,recognitionisnaturalratherthanexperimen- ologicalarousal. ThePCSmodelassumesthatinforma- tallyinduced,anditishighlycorrelatedwiththecriterion tionintegrationisbasedonautomatic-intuitiveprocesses (Pachur et al., 2008). The selection of this task ensures, JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 287 accordingtoproponentsoftheRH,thatparticipantswill theotheroneisnot(e.g.,shampooB),butwhenlooking use the RH and allows us to focus on the processes that at them on a supermarket shelf additional probabilistic underlietheadoptionofthisheuristic. cuesbecomeimmediatelyavailable. Suchcuesmightin- Importantly,inordertothoroughlyinvestigatethecog- cluderecommendationsofproducttestingagencies,eco- nitive processes underlying the use of recognition infor- labels,sizeandcolorofthepackage,etc. Thus,inmany mation, two major methodological considerations must real-lifesituations,accountingforexternalinformationis betakenintoaccount. First,choicepredictionsofdiffer- inherentintheevaluationofrecognitioninformation. In ent models may overlap, and thus multiple means of in- thecurrentpaperweaimedtocapturethesekindsofsit- vestigationareinorder(Ayal&Hochman,2009;Glöck- uations,andtoshedlightontheprocessesunderlyingthe ner, 2009, Glöckner & Bröder, in press; for a choice- evaluationofrecognitionandadditionallyavailableinfor- basedapproachseealsoMarewskietal.,inpress). Here mation. we look at choice patterns, response times, and confi- dencelevel. Inaddition,wefocusonautonomicnervous 1.2 Indirect measures of processes and hy- systemarousal,whichisknowntoaccompanyemotional potheses responsestopsychologicalstimuli(Andreassi,2000;An- noni, Ptak, Caldara-Schnetzer, Khateb, & Pollermann, 1.2.1 Choiceproportion 2003; Cacioppo, Tassinary, & Berntson, 2007), serving Choiceproportionreferstotheproportionofindividuals’ asaphysiologicalmarkerofbehavioralresponses(Sokol- choicesthatareconsistentwiththepredictionsofaspe- Hessneretal.,2009). Tomeasureemotionalarousal,we cificdecisionmodel(alsoknownasadherencerates). If usedPeripheralArterialTone(PAT)atthefingertip. For choice tasks are diagnostic (i.e., if the different models further information on the PAT measure, see Hochman, predictdifferentchoices),thenthismeasurecanbeused Glöckner,andYechiam(2010). asanindicatoroftheunderlyingprocessesthatleadtothe Second,tocontrastpredictionsfromdifferentmodels, decision(Bröder&Schiffer,2003;foradiscussionofthe asufficientnumberofdiagnosticcases(inwhichthecon- problemsthatmightresultfromusingchoiceproportions sideredmodelshavecontradictingpredictions)shouldbe aspredictorforstrategyuseseealsoHilbig,inpress). included (Glöckner & Betsch, 2008a). In the current According to the RH, recognition information is used study, several trials will include additional information inanoncompensatorymanner. Thus,ifamajorityofpar- (in the form of dichotomous cues), which will be ei- ticipantsappliesRHitwouldbepredictedthatahighpro- ther congruent or incongruent with the naturally avail- portionof choices followsrecognition in cases in which ablerecognitioninformation(pointinginthesameorthe it is diagnostic (i.e., one option is recognized and the opposite direction, respectively).2 By focusing on trials other is not). In addition, this choice proportion should inwhichonlyrecognitioninformationisavailableversus be insensitive to the presence of additional information, trialsinwhichadditionalcongruentorincongruentinfor- whether this additional information is congruent or in- mation is presented, we were able to create a sufficient congruent with recognition. In contrast, the PCS model numberofsuchdiagnosticcases. (as well as other compensatory models) assumes that Thus, our methodological design enables us to juxta- allavailableinformationisintegratedinacompensatory pose contradicting predictions about the compensatory manner. Thus, thePCSmodelpredictshighproportions andnoncompensatoryuseofrecognitioninformation.By ofchoiceinlinewithrecognitionwhenrecognitioninfor- providingfurtherinformationaboutthecities,wedeviate mation and the additional cue are congruent. However, from the original domain of choices for which the RH when only recognition information is available, choice was suggested, namely to account for inferences from proportionwillbelower,andlowestyetifrecognitionin- memory.Note,however,thatinferencesthatincludeboth formation and the additional cue are incongruent. Since given information and information retrieved from mem- previousresearchsupportsacompensatoryintegrationof ory are rather common and highly relevant in everyday recognition information and further cues (i.e., the PCS life (for a similar approach see Glöckner & Bröder, in hypothesis),weexpectedtoreplicatethesefindings. press). Consider, for example, a consumer decision to buyashampoo.Oneproductisknown(e.g.,shampooA), 1.2.2 Responsetime 2Itshouldbenotedthattheuseofinducedknowledgemayhighlight Response time refers to the time it takes the decision theadditionalinformationandthuscanpotentiallyincreaseitsusein thedecisionprocess(seePachuretal., 2008). Toallowforastrong makertocontemplatebeforemakingherdecision. Short testandtominimizethepotentialimpactofdemandeffects,theaddi- decision latencies are assumed to reflect short process- tionalcueinformationhadlowvalidity(whichwasmuchlowerthanthe ingtimeandrelianceoninstantlyaccessibleinformation, validityofrecognitioninformation). Furthermore,wetriedtosupport whilelongdecisionlatenciesreflectmoredeliberatecog- participants’comprehensionbyexplainingthatthevalidityofthecues isonlysomewhathigherthanchancelevel. nitive consideration (Ayal & Hochman, 2009; Betsch, JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 288 2008;Horstmann,Ahlgrimm,&Glöckner,2009;Payne, ableinformationandontheresultsofintegrationofvar- Bettman,&Johnson,1993). ious cues (e.g., Ayal & Hochman, 2009; Erev, Wallsten, Since the recognition information used in the current & Budescu, 1994). Specifically, confidence level is as- study is natural, it is assumed to be instantly accessible sumedtodependontheperceiveddifferencebetweenthe (Pachur & Hertwig, 2006). Thus, the RH (in the imple- alternatives (Glöckner & Betsch, 2008c). Accordingly, mentation suggested by Gigerenzer & Goldstein, 1996) thePCSmodelpredictsthatconfidencelevelwillbehigh- predicts short response times on those cases in which estwhenallinformationpointstowardselectingthesame the recognition information is diagnostic. In addition, option,andlowestwhentherecognitioninformationand sincethisnoncompensatorymodelpredictsthatrecogni- theadditionalinformationareincongruent.Whenlessin- tion is the only information to be considered, this short formationisavailable(e.g.,onlyrecognition),confidence responsetimeshouldnotbeaffectedbyadditional(con- level is predicted to be intermediate. Based on findings gruent and incongruent) cue information. By contrast, in probabilistic inference tasks (Glöckner & Hodges, in the PCS model suggests that all available information press)andrecentfindingsinrecognitionbasedprobabilis- is automatically integrated to form a mental representa- ticinferences(Glöckner&Bröder,inpress),weexpectto tion. Responsetimeshouldbeshorterwhenconsistency findsupportforthePCShypothesiswiththismeasureas ishigh,becauseonealternativeismoreclearlysupported well. and thus selected more quickly (for a similar prediction see Bergert & Nosofsky, 2007). Thus, response time 1.2.4 Physiologicalarousal shouldbeshortestwhenbothavailablepiecesofinforma- tion (i.e., recognition and cue) are congruent. Response Thesympatheticbranchoftheautonomicnervoussystem time should increase when only recognition information is responsible for mobilizing bodily resources to allow is available (since the advantage of the preferred alter- theregulationandcoordinationofemotionalresponsesto nativeislessobvious). Finally, responsetimeshouldbe psychologicalstimuli(Andreassi, 2000). PATatthefin- longestwhenincongruentinformationispresented. Note gertips represents changes in blood volume in response thatHilbigandPohl(2009)haverecentlyprovidedcon- to environmental stimuli (Andreassi, 2000), and is con- vincingevidenceforthedecisiontimepredictionsofthe sidered a good measure of sympathetic autonomic sys- PCS models (and decision field theory; Busemeyer & temactivation(Lavie,Schnall,Sheffy,&Shlitner,2000). Townsend, 1993) and we expect to replicate their find- Thelevelofsympatheticactivationisindicatedbyade- ings. crease in the transparency of the blood as a function of bloodpressure. Foreachheartbeat,bloodpressurevaries betweensystolicanddiastolicpressures. Thesystolicar- 1.2.3 Confidencelevel terialpressureisdefinedasthepeakpressureinthearter- The confidence level reflects how confident a decision ies,whichoccursnearthebeginningofthecardiaccycle makerisaboutherinference. Accordingtothenoncom- (whentheheartconstricts);thediastolicarterialpressure pensatory principles, decisions are based solely on the isthelowestpressure(attherestingphaseofthecardiac first diagnostic piece of information (e.g., Gigerenzer et cycle) (Klabunde, 2005). PAT is the difference between al.,1999).Assumingthatcuesareinvestigatedinorderof the maximum and minimum pressures measured. PAT theirvalidity,decisionsshouldbebasedonthemostvalid has been successfully validated by showing high corre- diagnosticcue. Recognitioninformationhashighnatural lations with cortical arousal (Penzel, Fricke, Jerrentrup, validityinseveralsituations(e.g., Goldstein&Gigeren- Peter,&Vogelmeier,2002). zer, 2002; Pachuretal., 2008), andisassumedtobethe Affective responses in humans are accompanied by first step in a number of fast and frugal strategies, such increased sympathetic activation, indexed by vasocon- as the take-the-best heuristic (Gigerenzer & Goldstein, striction, i.e., a decreased PAT signal. Thus, autonomic 1996). Thus, decisionsthatarebasedonrecognitionin- arousal, as indexed by PAT, can be used to examine the formation alone should lead to a high confidence level wayinwhichrecognitioninformationisbeingtreatedand whichisassumedtobeequaltothevalidityoftherecog- integratedduringthedecisionprocess. nition cue (Gigerenzer, Hoffrage, & Kleinbölting, 1991; In particular, if recognition represents a natural affec- seealsoAyal&Hochman,2009). Furthermore,sincethe tivesignal, thenweshouldexpectanincreaseinarousal processinthesediagnosticrecognitioncasesisassumed inresponsetorecognizedstimulirelativetounrecognized tobenoncompensatory(i.e.,basedonlyonthediscrimi- information.Regardlessofwhetherrecognitionactsasan natinginformationofrecognition),thishighlevelofcon- affective signal, according to the noncompensatory per- fidenceshouldnotbeaffectedbyadditionalinformation. spective,whenrecognitiondifferentiatesbetweentwoal- On the other hand, compensatory principles suggest ternatives,nochangesinarousalshouldbeexpectedwith thatconfidenceestimationsshouldbebasedonallavail- the presentation of additional information, whether this JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 289 Table1: CompetingpredictionsoftheRHandPCSmodelsforeachofthemeasuresusedinthecurrentstudy. Measure RHpredictions PCSpredictions Choice Choiceproportionalignedwiththe Choiceproportionsfollowingrecognitionwillbe proportion recognizedoptionwillbehigh,andwill highestforadditionalcongruentinformation,and notbeaffectedbyadditional lowestforincongruentinformation. Intermediate information,whetherthisinformationis choiceproportionsareexpectedwhenonly congruentorincongruentwith recognitioninformationisavailable. recognitioninformation. Responsetime Responsetimewillbefastifrecognition Responsetimewillbeshortestforadditional informationclearlyfavorsonechoice informationcongruentwithrecognition overanother,andshouldnotbeaffected informationandlongestforincongruent byadditionalinformation,whether information. Intermediateresponsetimeis congruentorincongruentwith expectedwhenonlyrecognitioninformationis recognitioninformation. available. Confidencelevel Confidencelevelwillbehighif Confidencelevelwillbehighestforadditional recognitioninformationisavailable,and informationcongruentwithrecognition shouldnotbeaffectedbyadditional informationandlowestforincongruent information,whethercongruentor information. Intermediateconfidencelevelis incongruentwithrecognition expectedwhenonlyrecognitioninformationis information. available. Physiological Recognizedcitynameswillbe Physiologicalarousalwillbelowerforcongruent arousal associatedwithincreasedphysiological informationrelativetoincongruentinformation, arousalrelativetounrecognizedcity reflectingincreasedsensitivitytolowconsistency names.∗Inaddition,physiological versushighconsistency. arousalwillnotbeaffectedby additionalinformation,whether congruentorincongruentwith recognitioninformation. ∗NotethatthispredictionisnotdirectlyderivedfromtheoriginalRH-specificationofGoldsteinandGigeren- zer (1999; 2002). Rather, it represents a proposition put forward in the current study to examine whether recognitioninformationcreatesanaffectivesignalfavoringoneoptionthatdominatesallotherinformation. information is congruent or incongruent with the recog- Predictions concerning all dependent variables are nitioninformation. summarizedinTable1. By contrast, if recognition takes no precedence over other cognitive cues, and is integrated with other infor- mation, then autonomic arousal should not depend on recognition alone. According to the PCS model, physi- 2 Method ologicalarousalshouldbedependentontheconsistency (vs.conflict)amongallavailablecueinformation(Glöck- ner&Hochman,inpress). Hence,itshouldbehighestin 2.1 Participants low consistency situations (i.e., when recognition infor- mationandtheadditionalcueareincongruent),andlow- Twenty-four undergraduates (10 females; mean age = estinhighconsistencysituations(i.e.,whenbothtypesof 24.0 years, SD = 2.1) from the Technion — Israel In- informationarecongruent). Thecurrentstudyisthefirst stitute of Technology served as paid participants in the to investigate the relation between recognition informa- experiment. They received a flat fee of 80 NIS (approx. tion and physiological arousal in probabilistic inference $21) and had a chance to get an extra bonus of 50 NIS tasks,andthisisthemaincontributionofthiswork. (approx. $13)dependingontheirperformance. JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 290 Table2: Citiesusedintheexperiment. 2.4 Designandprocedure Recognizedcities Unrecognizedcities Participants were presented with the 190 pairs of cities, Sacramento Raleigh and were asked to choose the city with the larger pop- Honolulu Plano ulation in each pair. No feedback was provided after Buffalo Greensboro choice. A confidence measure appeared right after each selection,andparticipantswererequiredtostatehowcer- NewOrleans BatonRouge taintheywereabouttheirchoice(onascalefrom50%to Orlando Amarillo 100%,in10%intervals).Reactiontimewasmeasuredby SaltLakeCity Topeka acomputerprogram,startingfromtheonsetofeachpair Hollywood Lansing ofcitiesuntilthechoiceresponse. AnnArbor Harrisburg In addition, in 91 of the trials (50 in the critical trials Cambridge JeffersonCity inwhichrecognitiondifferentiated),acueintheformof Annapolis Juneau a+/-signwaspresented. Thissignconstitutedthemain within-participantsmanipulation. Onaverage, inhalfof thecriticaltrials(i.e.,thecongruentcuetrials)thecueled 2.2 Materials to choosing the recognized city (i.e., a plus sign point- ing towards the recognized city or minus sign pointing Apilotstudywasconductedtocreateappropriatemateri- als. Inthepilotstudy,weassessedtherecognitionrateof toward the unrecognized city) and hence was congruent 48UScitiesbyIsraeliparticipants(N=20).The10most with the recognition cue. In the other half of the trials recognized cities (> 75%) and the 10 least recognized (i.e., the incongruent cue trials) the cue pointed towards cities (< 15%) were chosen (see Table 2). A complete selectingtheunrecognizedcity.Hence,theadditionalcue pair-wisecombinationofcitynamesresultedinasample was presented for recognized and unrecognized cities. of190pairsthatwereusedintheexperiment. Onehun- Participants were informed that this cue represented the dred trials were used as the critical trials, in which we prediction of a geography student, who makes a correct expected that participants would recognize one city, but prediction about the city with the larger population in nottheother(althoughtheanalysistookintoaccountac- 6 out of 10 cases. This reported validity was obtained tualrecognitionratesforeachparticipant). Theother90 in the experimental manipulation; namely the additional trialswereusedasfillertrialsthatweredesignedtomask cuewascorrectin60%ofthetrials. Toensurethatpar- thepurposeoftheexperiment(i.e.,trialsinwhichrecog- ticipantsrealizedthatthecuesignifiedlowvalidity,they nition does not differentiate between options). Because werefurtherinformedthatavalidityof50%inthistask wewereinterestedintheinfluenceoftherecognitioncue (i.e., 5 out of 10 correct) is equivalent to chance level. alone, weselectedcitiessothattwoadditionalcuesthat Half of the cues presented a minus sign, and the other mightbe potentiallyused were balanced. That is, inthe half presented a plus sign. Similarly, half of the minus recognizedandunrecognizedsamplethenumberofstate signs were adjacent to the recognized city and the other capital cities and the number of cities with universities were equal. In addition, all cities had medium to small halfwasassociatedwiththeothercity,andthesamewas population(rangingfrom30,000to460,000),inorderto true for the plus signs. Each participant saw all trials, avoidceilingeffectsoffamiliarity. eachinadifferentrandom order. Tomotivategoodper- formance,abonusof50NISwasofferedtothosewhose successratewashigherthan75%(acrossalltrials). 2.3 PATdataacquisition Importantly,the20citieswerepresentedseparatelyin PAT data were obtained using the SitePAT-200 (Ita- a different test, and for each city participants were re- mar Medical Ltd., Keisaria, Israel), a photo-cell sensor quired to state whether they recognize it or not. Half of plethysmograph,shapedasafingercup,whichisplaced the participants were given this recognition test prior to attheendofthefirstfingerofthenon-dominanthand(see thecity-sizetask,andtheotherhalfreceiveditaftercom- Karasik et al., 2002). The participant’s non-dominant hand was fixed on a hand rest during the whole ses- pletion of the city-size task. Since order turned out to sion. Therateofdataacquisitionwas100Hz, averaged have no effect, the data were collapsed across these ex- to 1 sample per second. Autonomic activity associated perimental conditions. The inter-trial interval was set to with arterial constriction (which indicates arousal) leads 7 seconds in both phases, so as to minimize residual ef- tolowervaluesonthePATmeasure(expressedbyvolts). fectsofpriorphysiologicalresponses. JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 291 RH makes no specific predictions about cases in which Figure 1: Mean choice proportions as a function of the bothcitiesarerecognizedorunrecognized,3RTandcon- typeofinformationavailablebesidesrecognition. fidencelevelshouldstillsystematicallydiffertocasesin 0 1. whichinspectingrecognitioninformationaloneinstantly leadstoadecision. Morespecifically, whenotherinfor- 9 mation than recognition has to be considered, we gen- n 0. o erally expect a slower reaction time and a lower level ecogniti 0.8 obfyctohnefifidnendcineg.sI.ndWeehde,nthreesceogpnreitdiiocntiownasswneortedsiuapgpnoorstteidc d r (i.e., both cities are known) or not available (i.e., both owe 0.7 cities are unknown), average response time across par- s foll ticipantswas9.03seconds(SD=0.93). Responsetimes oice 0.6 dropped to 8.48 seconds (SD = 0.88) when recognition h information differentiated between options. A within- c % 5 subjectrepeatedmeasuresANOVArevealedthatthisdif- 0. ference was significant [F (1, 23) = 15.781, p = 0.001]. Asimilarpatternofresultswasfoundforconfidencerat- 4 0. ings.Namely,whenrecognitionwasnotdiagnosticorun- No cue Congruent cue Incongruent cue available, the average confidence level was 60.42% (SD = 6.4). Confidence level increased to 66.3% (SD = 9.4) 3 Results whenrecognitioninformationdifferentiatedbetweenop- tions. A within-subject repeated measures ANOVA re- vealed that this difference was significant [F (1, 23) = 3.1 Recognitionandcuevalidityestimates 19.178,p<0.0001]. On average, participants recognized approximately 9 Taken together, these results support the first part of cities (SD = 1.5; range 5–12). A one-way analysis of theRHhypothesesaccordingtowhichdiagnosticrecog- variance (ANOVA) revealed that the order of this task nition information leads to high choice proportions that (i.e., beforeoraftertheexperimentaltask)hadnoeffect fit the recognition cue, short RT, and a high confidence onrecognitionranks[F(1,23)=0.259,p=0.616]. This level (see Table 1). The results also support the validity resultsuggeststhatparticipantswereabletodifferentiate ofourmeasuresinthistask. theirexperience-basedrecognitionfrommerefamiliarity with the city names that was induced by repeated expo- 3.2 Choiceproportionanalysis suretothecitynamesintheexperiment. The validity of the recognition cue across all critical The results of the choice proportion analysis are pre- trials in which only one city was recognized was 77% sented in Figure 1. As mentioned above, participants’ (SD=12.6%),andsignificantlyabovechancelevel[t(23) choices followed recognition in, on average, 83.3% of = 10.724, p < 0.0001 in a one-sample t-test]. This anal- thecasesinwhichnoinformationotherthanrecognition ysis confirms the ecological rationality of using the RH was presented. This proportion increased to 95.1% (SD in the current study, establishing the task as a fair test = 6.8%) when a cue congruent with the recognition in- forexaminingtheprocessesunderlyingtheprocessingof formationwasintroducedanddecreasedto50.1%(SD= recognition information. Importantly, the validity of the 26%)whenanincongruentcuewaspresented. Awithin- recognitioncuewasfarabovethe60%validityofthead- subjectrepeatedmeasuresANOVAwasconductedtotest ditional cognitive cue that was provided in some of the the effect of cue type (i.e., no cue, congruent cue, and trials(i.e.,informationfromthegeographystudent). incongruent cue) as the independent variable on choices Next,weexaminedwhethertheuseofrecognitionin- which followed recognition as the dependent variable. formationaffectedchoicebehavior. Todoso, wecalcu- This analysis revealed a significant main effect of cue latedtheproportionofchoicesthatfollowedtherecogni- type [(F (2, 69) = 59.26, p < 0.0001). Post-hoc paired- tioncue,andcomparedittochancelevel.Aone-sampled sample analyses demonstrated that the no-cue choices t-test showed that choices were sensitive to the recogni- differedsignificantlyfromboththecongruentandthein- tion cue in the majority of cases in which only recog- congruent cue choices [t(23) = -4.919, p < 0.0001, and nition information was available [average = 83.3%; SD t(23)=7.265,p<0.0001,respectively]. = 13%; t(23) = 12.889, p < 0.0001]. In addition, we examined whether response times and confidence level 3SincenodifferencewasfoundbetweentheRTsforeitherrecog- were affected by recognition information. Although the nizedornone-recognizedcitypairs,theseresultswerecollapsed. JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 292 on only one piece of information in a noncompensatory Figure2:A-Meanresponsetimeasafunctionofthetype manner. Instead, theresultssuggestthatparticipantsare of information available besides recognition. B - Mean sensitive to all information, and base their decisions on confidencelevelasafunctionofthetypeofinformation itsintegration. available besides recognition. Error bars represent stan- Partially in contrast to previous findings (e.g., Gold- darderrors. stein & Gigerenzer, 2002; Pachur et al., 2008), it seems 5 9. A that recognition information was completely overridden by other information, even when the additional cue was of low validity and was obtained from a relatively un- 0 known source. In fact, when a cue — with a validity as c) 9. lowas.60—thatwasincongruentwithrecognitionwas e s e ( presented,iteliminatedtheeffectoftherecognitioncue, m so that the proportion of choices following recognition se ti 8.5 droppedtochancelevel. Thus,inlinewithmostpriorre- n po search,thecurrentresultssuggestthatrecognitioninfor- s Re mation is only one of several cues that are integrated in 0 acompensatorymanner. Moreimportantly, ourfindings 8. suggest that real experience-based recognition informa- tionisnotsignificantlystrongerthanevenalowvalidity cognitivecue. Finally,assuggestedbypreviousresearch(e.g.,Pachur No cue Congruent cue Incongruent cue et al., 2008), individual differences might exists in peo- 2 ple’suseofrecognitioninformationandadditionalcues. 7 Toexaminethisalternativespeculationwecalculated,for B each individual, the correlation coefficient between pro- 0 7 portionsofchoicesthatfollowedrecognitionandcondi- tion (i.e., congruent cue, no cue, and incongruent cue). %) el ( 68 The median correlation coefficient between choice and v conditionacrossparticipantswas–0.95(t(23)=–16.406, e e l p < 0.05). Similarly, for 95.6% of the participants the enc 66 correlationturnedouttobehighlysignificant(p<0.05). d nfi Thispatternofindividualresultssupportstheaggregated o C 4 results and suggests that choice proportions were high- 6 estwhenacongruentcuewaspresentedandlowestwhen an incongruent cue was presented even at the individual 2 6 level. Thereby,individualdifferencescannoteasilyserve as alternative explanation for our results, as there seems No cue Congruent cue Incongruent cue tobeonlysmallindividualdifferencesintheuseofrecog- nitioninformationandadditionalcues. These results replicate previous findings (e.g., Bröder 3.3 Responsetimeanalysis & Eichler, 2006; Oppenheimer, 2003; Richter & Späth, 2006) and support the qualitative predictions concern- TheanalysisofresponsetimesispresentedinFigure2A. ing changes in choice proportion derived from the PCS Overall,themeanresponsetimeforcasesinwhichonly model. Theresultsconflictwithpredictionsderivedfrom one city was recognized was 8.48 seconds (SD = 0.88) theRH.Specifically, theresultssuggestthatrecognition whennootherinformationwaspresented.Responsetime information is integrated with additional information in decreased to 8.11 seconds (SD = 0.84) when a congru- a compensatory manner. The claim that recognition in- ent cue was introduced and increased to 9.14 seconds formation is used as the only information in a noncom- (SD = 1.32) when an incongruent cue was presented. pensatorymannerhastoberejected. Moreover,thispat- A within-subject repeated measures ANOVA was con- tern of results does not even support a weaker interpre- ducted to test the effect of cue type (i.e., no cue, con- tation of the RH (Brandstätter, Gigerenzer, & Hertwig, gruent cue, and incongruent cue) on response time as 2008),whichassumesthat,whiledecisionmakersevalu- the dependent variable. This analysis revealed a signif- atealltheavailableinformation,thefinalchoiceisbased icant main effect of cue type [(F (2, 69) = 17.472, p < JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 293 0.0001). Post-hoc paired-sample analyses demonstrated 9.4%)whennootherinformationwaspresented. Confi- thatresponsetimeswithoutanycuedifferedsignificantly dencelevelincreasedto69.4%(SD=10.1%)whenacon- fromresponsetimesfollowingboththecongruentandthe gruentcuewasintroducedanddecreasedto64.3%(SD= incongruent cue [t(23) = 2.692, p = 0.01, and t(23) = – 7.9%)whenanincongruentcuewaspresented.Awithin- 3.136,p=0.005,respectively]. subjectrepeatedmeasuresANOVAwasconductedtotest Thus,similartothechoiceproportionresults,decision the effect of cue type (i.e., no cue, congruent cue, and time results support the predictions of the PCS model, incongruent cue) on confidence level as the dependent replicating findings by Hilbig and Pohl (2009). In line variable. Thisanalysisrevealedasignificantmaineffect with the predictions of the PCS model, our findings of cue type [(F (2, 69) = 25.361, p < 0.0001). Post-hoc demonstrate that when cue predictions are incongruent paired-sampleanalysesdemonstratedthattheconfidence (i.e., when recognition information points at one choice levelintheno-cuechoicesdifferedsignificantlyfromthe andthecuepointsattheoppositechoice),responsetime confidencelevelofboththecongruentandtheincongru- isincreased. Furthermore,wefoundsupportforthePCS entcueconditions[t(23)=-4.791,p<0.0001,andt(23) hypothesis that decision times are decreased if the addi- =2.73,p=0.01,respectively]. tional cognitive cue points in the same direction as the Again, the results support the predictions of the PCS recognitioncue.Thisfindingisimportantinshowingthat model but not the predictions derived from the RH. In more information (i.e., the recognition cue + the cogni- particular,ourfindingssuggestthatavailableinformation tive cue) is processed faster than less information (i.e., isintegratedandthatitaffectsconfidencelevel.Theanal- recognitioncuealone). Theseeffectsspeakagainstserial ysis of confidence level shows that rather than relying modelsofinformationintegration(e.g.,Payne,Bettman, solely on recognition information, decision makers con- &Johnson,1988;seealsoGlöckner&Betsch,2010). siderrecognitiontobejustoneofothercuesthatareinte- Moreover, ifrecognitioninformationwereconsidered gratedbyacompensatoryprocessonthewaytoreaching in a noncompensatory manner, when no additional cue adecision. was presented, response times should have been equal At the individual level, the median correlation coeffi- to the other two conditions (either congruent or incon- cientbetweenconditionandconfidencelevelacrosspar- gruent cues) or even shorter. Shorter response times for ticipants was –0.89 (t(23) = –13.599, p < 0.0001), sug- recognition alone could support a weaker interpretation gesting that confidence level was highest for congruent ofthenoncompensatoryviewpoint,inwhichallavailable trialsandlowestforincongruenttrials.Similarly,for92% informationisscannedbutthefinaldecisionisbasedon oftheparticipantsthiscorrelationturnedouttobehighly thebestcuealone. Insuchcases,decisionmakerswould significant(p<0.0001). Thus, thisresultaddsuptothe havelessinformationtoevaluateandcouldthusbefaster. choiceproportionsandresponsetimesfindings, andfur- However,theresultsprovidenosupportforthisweakin- thersupporttheclaimthatthecurrentresultsdonotstem terpretationofthenoncompensatorymodels. Rather,the from individual differences in the use of recognition in- patternofresultsareinlinewithPCSmodels,suggesting formation. that more processing time is required in cases in which itishardertofindaconsistentmentalrepresentationdue 3.5 Physiologicalarousal tohighconflict(i.e.,lowconsistency)betweenpiecesof information. 3.5.1 Emotionalresponsetorecognizedstimuli Finally,asforchoiceproportions,wecalculated,atthe individual level, the correlation coefficient between re- To evaluate whether recognition information creates an sponse time and condition. The median correlation co- affective signal, we compared physiological reaction, efficientbetweenresponsetimesandconditionwas0.93 measuredforthetaskinwhichparticipantsratedwhether (t(23)=9.235,p<0.05),suggestingthatattheindividual theyrecognizedthecityornot(forsinglecities). Hence, level response times were highest for incongruent trials wemeasuredphysiologicalreactioninresponsetorecog- and shortest for congruent trials. Similarly, for 86% of nized city names versus unrecognized city names. Note theparticipants,thiscorrelationwashighlysignificant(p thatlowPATmeasuresindicatehigharousal. Ascanbe < 0.05). Thus, as in choice proportions, this result ex- seen in Figure 3A, there was a slight tendency that rec- cludesanindividualdifferencesaccountofourdata. ognized cities led to higher arousal, particularly 2 to 4 secondsafterstimulusonset.However,thistendencywas 3.4 Confidencelevelanalysis notstatisticallysignificant. From2secondspriortocity name presentation to 7 seconds after presentation, there Theresultsoftheconfidencelevelanalysisarepresented was no significant difference between PAT responses to inFigure2B.Overall,themeanconfidencelevelforcases the different types of stimuli. Namely, the average al- inwhichonlyonecitywasrecognizedwas66.3%(SD= phaacrossallwithin-subject(recognizedversusunrecog- JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 294 0.91 in this analysis, making it unlikely (i.e., beta-error Figure 3: A — Average PAT scores in response to rec- = 0.09) that an existing effect was merely not detected ognized versus unrecognized city names. B — Average (Faul,Erdfelder,Lang,&Buchner,2007). PAT scores as a function of information type. Low PAT Althoughnullhypothesisresultsshouldbeinterpreted scoresindicatehigharousal. cautiously, our findings imply that in the current task recognitioninformationdoesnotseemtocreateanaffec- 8 A tivesignalfavoringtherecognizedcity,oritmaycreatea 7. veryweaksignalwhichishardtodetect.Thisconclusion shouldbefurthervalidatedinfutureresearch. Unrecognized 6 7. Recognized s) olt 3.5.2 Physiological response to additional informa- v AT ( 7.4 tion P To examine physiological response to additional cue in- 2 formation, we compared PAT scores in response to con- 7. gruent versus incongruent trials.5 As mentioned above, PAT scores represent vasoconstriction, and low scores 7.0 thereforeindicateincreasedarousal. Ascanbeseenfrom Figure 3B, arousal was higher in incongruent trials than −2 −1 0 1 2 3 4 5 6 7 in congruent trials (i.e., lower PAT scores). A within- Time (relative to city name presentation subjectrepeatedmeasuresANOVArevealedthatthisdif- in the recognition task) ference was significant at one second after the presenta- tionofinformation[F(1,23)=4.09,p=0.05]. 2 B Thus, in line with the PCS model prediction, arousal 8. Incongruent increased(i.e.,lowerPATscoreswereobtained)withde- 0 Congruent creasing consistency between the available information. 8. Thearousaldatafurthersuggestthatadditionalinforma- tion is integrated rather than neglected, as would have olts) 7.8 been predicted by the RH. Specifically, the increased v PAT ( 7.6 acroonussisatlenincyr)e,sbpuotnnseottoincroenspfloicntsinegtoinrefocromgnaittiioonn(iin.efo.,rmloaw- tion alone, suggests that information is integrated in a 4 7. compensatorymannerandthatrecognitioncuesareused asoneofseveral(cognitive)cues. 2 7. 4 Discussion −2 −1 0 1 2 3 4 5 6 7 Time (relative to pair of cities and additional cue presentation in the city−size task) Thecurrentstudyaimedtoexaminethenatureofrecog- nition information in human inference tasks, as well as the integration of such information with additional cog- nized)repeatedmeasuresANOVAwas0.51(SD=0.25, nitive cues. In particular, we examined two main re- range0.18–0.99).4 Wecannot,however,ruleoutthepos- search questions. First, we tested inferences regarding sibility that the results might be partially due to the rel- ativelysmallsizeofoursample. Still, apost-hocpower 5ThePATscoreishighlysensitivetocognitiveload(Iani,Gopher, analysis(for10citieswithineachgroup,withanassumed Grunwald,&Lavie,2007;Iani,Gopher,&Lavie,2004),whichisas- correlationofr=0.3betweenobservationsandassuming sumedtoincreaseasthetaskprogresses. Tocontrolforthiseffect,we asmalleffectoff=0.25)indicatedanexcellentpowerof matchedforeachparticipantthenumberoftrialsineachcondition(con- gruentandincongruent),sothateachconditionhadthesamenumberof trialsastheconditionwiththeminimalnumberoftrials(Mean=17 4TheresultsoftherepeatedmeasureANOVAforthedifferencebe- trials).Itshouldbenoted,however,thatwhenweregressedoutthetime tweenthePATscoreforrecognizedvs.unrecognizedcities2to5sec- trend(i.e.,totaltrials)fromthePATscoresandusedregressionanalysis ondsafterstimuluspresentationswereF2sec(1,23)=1.941,p=0.18; topredictthePATscorefromthetypeofcue(congruentversusincon- F3sec (1, 23)=0.933, p=0.34; F4sec (1, 23)=0.727, p=0.40; and gruent)wefoundsimilarpatternofresults,althoughtheeffectofcue F5sec(1,23)=0.741,p=0.40. (Wedidnotcalculatethetestat1sec typewasonlymarginallysignificant(i.e.,thep-valueofthemodelwas becausetherewasnodifference,seeFigure3A). 0.055;one-tailedtest).

Description:
ing confidence ratings, response times, and choice proportions further demonstrated that recognition information and other cognitive cues are integrated in a compensatory manner. Keywords: affect, recognition heuristic, physiological arousal, parallel constraint satisfaction, noncompensatory models
ebook img

Physiological arousal in processing recognition information: Ignoring or integrating cognitive cues? PDF

15 Pages·2010·0.15 MB·English
by  
Save to my drive
Quick download
Download
Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Physiological arousal in processing recognition information: Ignoring or integrating cognitive cues?

JudgmentandDecisionMaking,Vol.5,No.4,July2010,pp.285–299 Physiological arousal in processing recognition information: Ignoring or integrating cognitive cues? GuyHochman∗ ShaharAyal Technion—IsraelInstituteofTechnology DukeUniversity AndreasGlöckner MaxPlanckInstituteforResearchonCollectiveGoods Abstract The recognition heuristic (RH; Goldstein & Gigerenzer, 2002) suggests that, when applicable, probabilistic infer- encesarebasedonanoncompensatoryexaminationofwhetheranobjectisrecognizedornot. Theoverallfindingson theprocessesthatunderliethisfastandfrugalheuristicaresomewhatmixed,andmanystudieshaveexpressedtheneed forconsideringamorecompensatoryintegrationofrecognitioninformation. Regardlessofthemechanisminvolved,it isclearthatrecognitionhasastronginfluenceonchoices,andthisfindingmightbeexplainedbythefactthatrecog- nitioncuesarouseaffectandthusreceivemoreattentionthancognitivecues. Totestthisassumption,weinvestigated whetherrecognitionresultsinadirectaffectivesignalbymeasuringphysiologicalarousal(i.e.,peripheralarterialtone) in the established city-size task. We found that recognition of cities does not directly result in increased physiolog- ical arousal. Moreover, the results show that physiological arousal increased with increasing inconsistency between recognitioninformationandadditionalcueinformation. Thesefindingssupportpredictionsderivedbyacompensatory ParallelConstraintSatisfactionmodelratherthanpredictionsofnoncompensatorymodels. Additionalresultsconcern- ingconfidenceratings, responsetimes, andchoiceproportionsfurtherdemonstratedthatrecognitioninformationand othercognitivecuesareintegratedinacompensatorymanner. Keywords:affect,recognitionheuristic,physiologicalarousal,parallelconstraintsatisfaction,noncompensatorymodels. 1 Introduction typical tools in this toolbox, suggests that probabilistic inferences about a specific criterion (e.g., which of two Imagine a business trip to a city you have never visited citiesismorepopulated),aremadeonthebasisofrecog- before. Your meetings end later than you had expected nitionalone,sothatrecognizedobjects(e.g.,thenameof andyoudecidetospendthenight.Youmakeafewphone thecity)willbechosenoverunrecognizedones. callsandfindtwohotelswithvacancies,bothatasimilar Indeed, it has been shown repeatedly that people ac- rate. Youknownothingelseaboutthesehotels;however, tuallypreferrecognizedobjectsoverunrecognizedones, youdorecognizethenameofoneofthem. Whichhotel and that such behavior may lead to rather accurate in- wouldyouchoose? AccordingtoGoldsteinandGigeren- ferences (e.g., Borges, Goldstein, Ortmann, & Gigeren- zer (1999; 2002), these kinds of probabilistic inferences zer, 1999; Goldstein & Gigerenzer, 1999, 2002; Pachur couldberesolvedbyrecognitioninformationalone. & Biele, 2007; Pohl, 2006; Reimer & Katsikopoulos, More specifically, according to Gigerenzer and col- 2004;Serwe&Frings,2006). Nevertheless,severalstud- leagues (Gigerenzer, 2000; Gigerenzer & Goldstein, ies challenge the claim that recognition information is 1996; Gigerenzer, Todd et al., 1999), individuals are used in a noncompensatory manner1 as predicted by the equipped with several tools (i.e., fast and frugal heuris- RH(e.g., Bröder&Eichler, 2006; Hilbig&Pohl, 2008; tics) that exploit the structure of the environment, some McCloy & Beaman, 2004; Newell & Fernandez, 2006; of which rely solely on one piece of information at a timeontheroadtoreachingadecision. Therecognition 1Noncompensatorystrategiessuggestthatdifferentcuesorattributes ofthechoicealternativeareconsideredinaspecificorder,usuallyin heuristic (RH), the simplest and one of the most proto- theorderoftheirvalidityorimportance(e.g.,thetakethebestheuristic, TTB).Onlyonepieceofinformationisbeingevaluatedatatimeand ∗Address:GuyHochman,MaxWertheimerMinervaCenterforCog- cuescannotcompensatefordeficitsinotherscues. Incontrast,com- nitive Studies, Faculty of Industrial Engineering and Management, pensatorystrategiesassumethatchoicesarebasedontheintegrationof TheTechnion-IsraelInstituteofTechnology, Haifa, Israel. E-mail: theavailableinformation.Thus,compensationbetweencuesispossible [email protected]. (e.g.,theExpectedUtilityTheory—EUT). 285 JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 286 Newell & Shanks, 2004; Oppenheimer, 2003; Pachur, —akintoperception(Holyoak&Simon,1999;McClel- Bröder&Marewski,2008; Richter&Späth, 2006). For land&Rumelhart,1981;Read,Vanman,&Miller,1997; example, in Oppenheimer (2003), Bröder and Eichler Thagard,1989)—bywhichconsistentinterpretationsof (2006), and Richter and Späth (2006), additional infor- decision tasks are constructed. Available information is mationhasbeenfoundtoaffecttheproportionofchoices taken into account according to its importance, and the thatfollowedrecognition. Whileithasbeenarguedthat advantages of the (emerging) preferred option are high- several of these studies suffer from methodological lim- lighted by systematic information distortions (see also itations(seePachuretal., 2008), theirfindingsnonethe- Russo, Carlson, Meloy, & Yong, 2008). Theformalized less highlight the need for considering a more compen- PCSmodelincludesfourpossiblesteps. First,whenpre- satoryintegrationofrecognitioninformation. sented with a decision task, related information is acti- Moreover, studies on the RH have focused mainly vatedtoformamentalrepresentationofthetask.Second, on choice behavior. Yet, since the choice predictions automaticprocessesofPCSleadtomaximizationofcon- made by different models often overlap (see Ayal & sistency by automatically highlighting information sup- Hochman, 2009; Glöckner & Betsch, 2008a), it is dif- portingthefavoredalternativeandatthesametimesup- ficult to draw clear conclusions concerning the underly- pressingcontradictinginformation. Asaresult,aconsis- ing cognitive processes. Several recent studies have at- tentrepresentationoftheavailableinformationisformed. tempted to address this problem, for example by testing Third, the decision maker evaluates the choice alterna- hypothesesconcerningprocessmeasuressuchasdecision tives, and if one alternative clearly dominates the other time(Hilbig&Pohl,2008;2009;Marewski,Gaissmaier, (i.e.,ifthereisenoughinformationinfavorofthisalter- Schooler, Goldstein, & Gigerenzer, 2010), and by addi- native), this alternative is chosen. However, if informa- tionallyinvestigatingconfidence(Glöckner&Bröder,in tionconsistencyfailstoreachacertainthreshold,thede- press). Yet, additional research is needed to disentangle cisionmakertakesaforthstep, inwhichdeliberatecon- theroleofrecognitioninformationindecisionmaking,as struction processes are activated to change the structure wellastoexaminethecognitiveprocessesunderlyingits ofthenetwork(e.g.,changingthepatternofinformation useininferencesandchoicebehavior. search,assigningdifferentweightstothedifferentpieces To achieve these goals, we investigated the nature of of information). Among other things, the PCS model is the processes underlying the use of recognition infor- supportedbyresearchthatshowsthatunconsciousmod- mation in probabilistic inferences, using physiological ificationsoftheavailableinformationoccurduringdeci- arousal, choice proportions, decision times, and confi- sion processes (e.g., Glöckner, Betsch, & Schindler, in dence as dependent measures. We derived concurring press; Holyoak&Simon, 1999; Simon, Krawczyk, Ble- predictions on the basis of a noncompensatory perspec- icher, & Holyoak, 2008; Simon, Snow, & Read, 2004). tive, and juxtapose these with contradicting predictions Furthermore, PCS models have been found to account derivedfromacompensatorymodel. Specifically,wein- well for both choices and decision times in probabilis- vestigate whether recognition leads to increased physio- ticinferencetasks(Glöckner,2008;Glöckner&Hodges, logical arousal, which dominates other cognitive infor- inpress;Glöckner&Bröder,inpress;seealsoGlöckner mation, as suggested by affect-based noncompensatory &Betsch,2008c). models(e.g.,Damasio,1994;Slovic,Finucane,Peters,& MacGregor, 2002). It has been argued that recognition 1.1 Overviewofthestudy information,likeotheraffectivesignals,mightbegener- ated automatically (Pachur & Hertwig, 2006), and that Thecurrentstudyisdesignedtoexaminewhetherrecog- itmayoverridecognitiveinformation(e.g., Goldstein& nition information is integrated with other information Gigerenzer,2002,Experiment2;Pachuretal.,2008,Ex- in a compensatory or noncompensatory manner, and periment1). Thus,weexaminedwhetherrecognitionin- whether recognition information is encoded as a special formationisaffectivelyencodedduringthedecisionpro- affective signal. To do so, we examined choice behav- cess. ior, while participants conducted a city-size task (Gold- We tested the RH against the compensatory Parallel stein & Gigerenzer, 2002; Richter & Späth, 2006). The Constraint Satisfaction (PCS) model for decision mak- city-size task was chosen in order to create an optimal ing (e.g., Glöckner & Betsch, 2008b). The PCS model environment for using the RH. Specifically, in this task was selected because it has been shown to account well thecriterion(i.e.,populationsize)isnotimmediatelyac- fordecisiontimes(Hilbig&Pohl,2009)andconfidence cessible, and inference has to be based on probabilistic (Glöckner & Bröder, in press) in recognition tasks, and cues(Goldstein&Gigerenzer,2002;Pachuretal.,2008). because it makes specific predictions concerning physi- Inaddition,recognitionisnaturalratherthanexperimen- ologicalarousal. ThePCSmodelassumesthatinforma- tallyinduced,anditishighlycorrelatedwiththecriterion tionintegrationisbasedonautomatic-intuitiveprocesses (Pachur et al., 2008). The selection of this task ensures, JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 287 accordingtoproponentsoftheRH,thatparticipantswill theotheroneisnot(e.g.,shampooB),butwhenlooking use the RH and allows us to focus on the processes that at them on a supermarket shelf additional probabilistic underlietheadoptionofthisheuristic. cuesbecomeimmediatelyavailable. Suchcuesmightin- Importantly,inordertothoroughlyinvestigatethecog- cluderecommendationsofproducttestingagencies,eco- nitive processes underlying the use of recognition infor- labels,sizeandcolorofthepackage,etc. Thus,inmany mation, two major methodological considerations must real-lifesituations,accountingforexternalinformationis betakenintoaccount. First,choicepredictionsofdiffer- inherentintheevaluationofrecognitioninformation. In ent models may overlap, and thus multiple means of in- thecurrentpaperweaimedtocapturethesekindsofsit- vestigationareinorder(Ayal&Hochman,2009;Glöck- uations,andtoshedlightontheprocessesunderlyingthe ner, 2009, Glöckner & Bröder, in press; for a choice- evaluationofrecognitionandadditionallyavailableinfor- basedapproachseealsoMarewskietal.,inpress). Here mation. we look at choice patterns, response times, and confi- dencelevel. Inaddition,wefocusonautonomicnervous 1.2 Indirect measures of processes and hy- systemarousal,whichisknowntoaccompanyemotional potheses responsestopsychologicalstimuli(Andreassi,2000;An- noni, Ptak, Caldara-Schnetzer, Khateb, & Pollermann, 1.2.1 Choiceproportion 2003; Cacioppo, Tassinary, & Berntson, 2007), serving Choiceproportionreferstotheproportionofindividuals’ asaphysiologicalmarkerofbehavioralresponses(Sokol- choicesthatareconsistentwiththepredictionsofaspe- Hessneretal.,2009). Tomeasureemotionalarousal,we cificdecisionmodel(alsoknownasadherencerates). If usedPeripheralArterialTone(PAT)atthefingertip. For choice tasks are diagnostic (i.e., if the different models further information on the PAT measure, see Hochman, predictdifferentchoices),thenthismeasurecanbeused Glöckner,andYechiam(2010). asanindicatoroftheunderlyingprocessesthatleadtothe Second,tocontrastpredictionsfromdifferentmodels, decision(Bröder&Schiffer,2003;foradiscussionofthe asufficientnumberofdiagnosticcases(inwhichthecon- problemsthatmightresultfromusingchoiceproportions sideredmodelshavecontradictingpredictions)shouldbe aspredictorforstrategyuseseealsoHilbig,inpress). included (Glöckner & Betsch, 2008a). In the current According to the RH, recognition information is used study, several trials will include additional information inanoncompensatorymanner. Thus,ifamajorityofpar- (in the form of dichotomous cues), which will be ei- ticipantsappliesRHitwouldbepredictedthatahighpro- ther congruent or incongruent with the naturally avail- portionof choices followsrecognition in cases in which ablerecognitioninformation(pointinginthesameorthe it is diagnostic (i.e., one option is recognized and the opposite direction, respectively).2 By focusing on trials other is not). In addition, this choice proportion should inwhichonlyrecognitioninformationisavailableversus be insensitive to the presence of additional information, trialsinwhichadditionalcongruentorincongruentinfor- whether this additional information is congruent or in- mation is presented, we were able to create a sufficient congruent with recognition. In contrast, the PCS model numberofsuchdiagnosticcases. (as well as other compensatory models) assumes that Thus, our methodological design enables us to juxta- allavailableinformationisintegratedinacompensatory pose contradicting predictions about the compensatory manner. Thus, thePCSmodelpredictshighproportions andnoncompensatoryuseofrecognitioninformation.By ofchoiceinlinewithrecognitionwhenrecognitioninfor- providingfurtherinformationaboutthecities,wedeviate mation and the additional cue are congruent. However, from the original domain of choices for which the RH when only recognition information is available, choice was suggested, namely to account for inferences from proportionwillbelower,andlowestyetifrecognitionin- memory.Note,however,thatinferencesthatincludeboth formation and the additional cue are incongruent. Since given information and information retrieved from mem- previousresearchsupportsacompensatoryintegrationof ory are rather common and highly relevant in everyday recognition information and further cues (i.e., the PCS life (for a similar approach see Glöckner & Bröder, in hypothesis),weexpectedtoreplicatethesefindings. press). Consider, for example, a consumer decision to buyashampoo.Oneproductisknown(e.g.,shampooA), 1.2.2 Responsetime 2Itshouldbenotedthattheuseofinducedknowledgemayhighlight Response time refers to the time it takes the decision theadditionalinformationandthuscanpotentiallyincreaseitsusein thedecisionprocess(seePachuretal., 2008). Toallowforastrong makertocontemplatebeforemakingherdecision. Short testandtominimizethepotentialimpactofdemandeffects,theaddi- decision latencies are assumed to reflect short process- tionalcueinformationhadlowvalidity(whichwasmuchlowerthanthe ingtimeandrelianceoninstantlyaccessibleinformation, validityofrecognitioninformation). Furthermore,wetriedtosupport whilelongdecisionlatenciesreflectmoredeliberatecog- participants’comprehensionbyexplainingthatthevalidityofthecues isonlysomewhathigherthanchancelevel. nitive consideration (Ayal & Hochman, 2009; Betsch, JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 288 2008;Horstmann,Ahlgrimm,&Glöckner,2009;Payne, ableinformationandontheresultsofintegrationofvar- Bettman,&Johnson,1993). ious cues (e.g., Ayal & Hochman, 2009; Erev, Wallsten, Since the recognition information used in the current & Budescu, 1994). Specifically, confidence level is as- study is natural, it is assumed to be instantly accessible sumedtodependontheperceiveddifferencebetweenthe (Pachur & Hertwig, 2006). Thus, the RH (in the imple- alternatives (Glöckner & Betsch, 2008c). Accordingly, mentation suggested by Gigerenzer & Goldstein, 1996) thePCSmodelpredictsthatconfidencelevelwillbehigh- predicts short response times on those cases in which estwhenallinformationpointstowardselectingthesame the recognition information is diagnostic. In addition, option,andlowestwhentherecognitioninformationand sincethisnoncompensatorymodelpredictsthatrecogni- theadditionalinformationareincongruent.Whenlessin- tion is the only information to be considered, this short formationisavailable(e.g.,onlyrecognition),confidence responsetimeshouldnotbeaffectedbyadditional(con- level is predicted to be intermediate. Based on findings gruent and incongruent) cue information. By contrast, in probabilistic inference tasks (Glöckner & Hodges, in the PCS model suggests that all available information press)andrecentfindingsinrecognitionbasedprobabilis- is automatically integrated to form a mental representa- ticinferences(Glöckner&Bröder,inpress),weexpectto tion. Responsetimeshouldbeshorterwhenconsistency findsupportforthePCShypothesiswiththismeasureas ishigh,becauseonealternativeismoreclearlysupported well. and thus selected more quickly (for a similar prediction see Bergert & Nosofsky, 2007). Thus, response time 1.2.4 Physiologicalarousal shouldbeshortestwhenbothavailablepiecesofinforma- tion (i.e., recognition and cue) are congruent. Response Thesympatheticbranchoftheautonomicnervoussystem time should increase when only recognition information is responsible for mobilizing bodily resources to allow is available (since the advantage of the preferred alter- theregulationandcoordinationofemotionalresponsesto nativeislessobvious). Finally, responsetimeshouldbe psychologicalstimuli(Andreassi, 2000). PATatthefin- longestwhenincongruentinformationispresented. Note gertips represents changes in blood volume in response thatHilbigandPohl(2009)haverecentlyprovidedcon- to environmental stimuli (Andreassi, 2000), and is con- vincingevidenceforthedecisiontimepredictionsofthe sidered a good measure of sympathetic autonomic sys- PCS models (and decision field theory; Busemeyer & temactivation(Lavie,Schnall,Sheffy,&Shlitner,2000). Townsend, 1993) and we expect to replicate their find- Thelevelofsympatheticactivationisindicatedbyade- ings. crease in the transparency of the blood as a function of bloodpressure. Foreachheartbeat,bloodpressurevaries betweensystolicanddiastolicpressures. Thesystolicar- 1.2.3 Confidencelevel terialpressureisdefinedasthepeakpressureinthearter- The confidence level reflects how confident a decision ies,whichoccursnearthebeginningofthecardiaccycle makerisaboutherinference. Accordingtothenoncom- (whentheheartconstricts);thediastolicarterialpressure pensatory principles, decisions are based solely on the isthelowestpressure(attherestingphaseofthecardiac first diagnostic piece of information (e.g., Gigerenzer et cycle) (Klabunde, 2005). PAT is the difference between al.,1999).Assumingthatcuesareinvestigatedinorderof the maximum and minimum pressures measured. PAT theirvalidity,decisionsshouldbebasedonthemostvalid has been successfully validated by showing high corre- diagnosticcue. Recognitioninformationhashighnatural lations with cortical arousal (Penzel, Fricke, Jerrentrup, validityinseveralsituations(e.g., Goldstein&Gigeren- Peter,&Vogelmeier,2002). zer, 2002; Pachuretal., 2008), andisassumedtobethe Affective responses in humans are accompanied by first step in a number of fast and frugal strategies, such increased sympathetic activation, indexed by vasocon- as the take-the-best heuristic (Gigerenzer & Goldstein, striction, i.e., a decreased PAT signal. Thus, autonomic 1996). Thus, decisionsthatarebasedonrecognitionin- arousal, as indexed by PAT, can be used to examine the formation alone should lead to a high confidence level wayinwhichrecognitioninformationisbeingtreatedand whichisassumedtobeequaltothevalidityoftherecog- integratedduringthedecisionprocess. nition cue (Gigerenzer, Hoffrage, & Kleinbölting, 1991; In particular, if recognition represents a natural affec- seealsoAyal&Hochman,2009). Furthermore,sincethe tivesignal, thenweshouldexpectanincreaseinarousal processinthesediagnosticrecognitioncasesisassumed inresponsetorecognizedstimulirelativetounrecognized tobenoncompensatory(i.e.,basedonlyonthediscrimi- information.Regardlessofwhetherrecognitionactsasan natinginformationofrecognition),thishighlevelofcon- affective signal, according to the noncompensatory per- fidenceshouldnotbeaffectedbyadditionalinformation. spective,whenrecognitiondifferentiatesbetweentwoal- On the other hand, compensatory principles suggest ternatives,nochangesinarousalshouldbeexpectedwith thatconfidenceestimationsshouldbebasedonallavail- the presentation of additional information, whether this JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 289 Table1: CompetingpredictionsoftheRHandPCSmodelsforeachofthemeasuresusedinthecurrentstudy. Measure RHpredictions PCSpredictions Choice Choiceproportionalignedwiththe Choiceproportionsfollowingrecognitionwillbe proportion recognizedoptionwillbehigh,andwill highestforadditionalcongruentinformation,and notbeaffectedbyadditional lowestforincongruentinformation. Intermediate information,whetherthisinformationis choiceproportionsareexpectedwhenonly congruentorincongruentwith recognitioninformationisavailable. recognitioninformation. Responsetime Responsetimewillbefastifrecognition Responsetimewillbeshortestforadditional informationclearlyfavorsonechoice informationcongruentwithrecognition overanother,andshouldnotbeaffected informationandlongestforincongruent byadditionalinformation,whether information. Intermediateresponsetimeis congruentorincongruentwith expectedwhenonlyrecognitioninformationis recognitioninformation. available. Confidencelevel Confidencelevelwillbehighif Confidencelevelwillbehighestforadditional recognitioninformationisavailable,and informationcongruentwithrecognition shouldnotbeaffectedbyadditional informationandlowestforincongruent information,whethercongruentor information. Intermediateconfidencelevelis incongruentwithrecognition expectedwhenonlyrecognitioninformationis information. available. Physiological Recognizedcitynameswillbe Physiologicalarousalwillbelowerforcongruent arousal associatedwithincreasedphysiological informationrelativetoincongruentinformation, arousalrelativetounrecognizedcity reflectingincreasedsensitivitytolowconsistency names.∗Inaddition,physiological versushighconsistency. arousalwillnotbeaffectedby additionalinformation,whether congruentorincongruentwith recognitioninformation. ∗NotethatthispredictionisnotdirectlyderivedfromtheoriginalRH-specificationofGoldsteinandGigeren- zer (1999; 2002). Rather, it represents a proposition put forward in the current study to examine whether recognitioninformationcreatesanaffectivesignalfavoringoneoptionthatdominatesallotherinformation. information is congruent or incongruent with the recog- Predictions concerning all dependent variables are nitioninformation. summarizedinTable1. By contrast, if recognition takes no precedence over other cognitive cues, and is integrated with other infor- mation, then autonomic arousal should not depend on recognition alone. According to the PCS model, physi- 2 Method ologicalarousalshouldbedependentontheconsistency (vs.conflict)amongallavailablecueinformation(Glöck- ner&Hochman,inpress). Hence,itshouldbehighestin 2.1 Participants low consistency situations (i.e., when recognition infor- mationandtheadditionalcueareincongruent),andlow- Twenty-four undergraduates (10 females; mean age = estinhighconsistencysituations(i.e.,whenbothtypesof 24.0 years, SD = 2.1) from the Technion — Israel In- informationarecongruent). Thecurrentstudyisthefirst stitute of Technology served as paid participants in the to investigate the relation between recognition informa- experiment. They received a flat fee of 80 NIS (approx. tion and physiological arousal in probabilistic inference $21) and had a chance to get an extra bonus of 50 NIS tasks,andthisisthemaincontributionofthiswork. (approx. $13)dependingontheirperformance. JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 290 Table2: Citiesusedintheexperiment. 2.4 Designandprocedure Recognizedcities Unrecognizedcities Participants were presented with the 190 pairs of cities, Sacramento Raleigh and were asked to choose the city with the larger pop- Honolulu Plano ulation in each pair. No feedback was provided after Buffalo Greensboro choice. A confidence measure appeared right after each selection,andparticipantswererequiredtostatehowcer- NewOrleans BatonRouge taintheywereabouttheirchoice(onascalefrom50%to Orlando Amarillo 100%,in10%intervals).Reactiontimewasmeasuredby SaltLakeCity Topeka acomputerprogram,startingfromtheonsetofeachpair Hollywood Lansing ofcitiesuntilthechoiceresponse. AnnArbor Harrisburg In addition, in 91 of the trials (50 in the critical trials Cambridge JeffersonCity inwhichrecognitiondifferentiated),acueintheformof Annapolis Juneau a+/-signwaspresented. Thissignconstitutedthemain within-participantsmanipulation. Onaverage, inhalfof thecriticaltrials(i.e.,thecongruentcuetrials)thecueled 2.2 Materials to choosing the recognized city (i.e., a plus sign point- ing towards the recognized city or minus sign pointing Apilotstudywasconductedtocreateappropriatemateri- als. Inthepilotstudy,weassessedtherecognitionrateof toward the unrecognized city) and hence was congruent 48UScitiesbyIsraeliparticipants(N=20).The10most with the recognition cue. In the other half of the trials recognized cities (> 75%) and the 10 least recognized (i.e., the incongruent cue trials) the cue pointed towards cities (< 15%) were chosen (see Table 2). A complete selectingtheunrecognizedcity.Hence,theadditionalcue pair-wisecombinationofcitynamesresultedinasample was presented for recognized and unrecognized cities. of190pairsthatwereusedintheexperiment. Onehun- Participants were informed that this cue represented the dred trials were used as the critical trials, in which we prediction of a geography student, who makes a correct expected that participants would recognize one city, but prediction about the city with the larger population in nottheother(althoughtheanalysistookintoaccountac- 6 out of 10 cases. This reported validity was obtained tualrecognitionratesforeachparticipant). Theother90 in the experimental manipulation; namely the additional trialswereusedasfillertrialsthatweredesignedtomask cuewascorrectin60%ofthetrials. Toensurethatpar- thepurposeoftheexperiment(i.e.,trialsinwhichrecog- ticipantsrealizedthatthecuesignifiedlowvalidity,they nition does not differentiate between options). Because werefurtherinformedthatavalidityof50%inthistask wewereinterestedintheinfluenceoftherecognitioncue (i.e., 5 out of 10 correct) is equivalent to chance level. alone, weselectedcitiessothattwoadditionalcuesthat Half of the cues presented a minus sign, and the other mightbe potentiallyused were balanced. That is, inthe half presented a plus sign. Similarly, half of the minus recognizedandunrecognizedsamplethenumberofstate signs were adjacent to the recognized city and the other capital cities and the number of cities with universities were equal. In addition, all cities had medium to small halfwasassociatedwiththeothercity,andthesamewas population(rangingfrom30,000to460,000),inorderto true for the plus signs. Each participant saw all trials, avoidceilingeffectsoffamiliarity. eachinadifferentrandom order. Tomotivategoodper- formance,abonusof50NISwasofferedtothosewhose successratewashigherthan75%(acrossalltrials). 2.3 PATdataacquisition Importantly,the20citieswerepresentedseparatelyin PAT data were obtained using the SitePAT-200 (Ita- a different test, and for each city participants were re- mar Medical Ltd., Keisaria, Israel), a photo-cell sensor quired to state whether they recognize it or not. Half of plethysmograph,shapedasafingercup,whichisplaced the participants were given this recognition test prior to attheendofthefirstfingerofthenon-dominanthand(see thecity-sizetask,andtheotherhalfreceiveditaftercom- Karasik et al., 2002). The participant’s non-dominant hand was fixed on a hand rest during the whole ses- pletion of the city-size task. Since order turned out to sion. Therateofdataacquisitionwas100Hz, averaged have no effect, the data were collapsed across these ex- to 1 sample per second. Autonomic activity associated perimental conditions. The inter-trial interval was set to with arterial constriction (which indicates arousal) leads 7 seconds in both phases, so as to minimize residual ef- tolowervaluesonthePATmeasure(expressedbyvolts). fectsofpriorphysiologicalresponses. JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 291 RH makes no specific predictions about cases in which Figure 1: Mean choice proportions as a function of the bothcitiesarerecognizedorunrecognized,3RTandcon- typeofinformationavailablebesidesrecognition. fidencelevelshouldstillsystematicallydiffertocasesin 0 1. whichinspectingrecognitioninformationaloneinstantly leadstoadecision. Morespecifically, whenotherinfor- 9 mation than recognition has to be considered, we gen- n 0. o erally expect a slower reaction time and a lower level ecogniti 0.8 obfyctohnefifidnendcineg.sI.ndWeehde,nthreesceogpnreitdiiocntiownasswneortedsiuapgpnoorstteidc d r (i.e., both cities are known) or not available (i.e., both owe 0.7 cities are unknown), average response time across par- s foll ticipantswas9.03seconds(SD=0.93). Responsetimes oice 0.6 dropped to 8.48 seconds (SD = 0.88) when recognition h information differentiated between options. A within- c % 5 subjectrepeatedmeasuresANOVArevealedthatthisdif- 0. ference was significant [F (1, 23) = 15.781, p = 0.001]. Asimilarpatternofresultswasfoundforconfidencerat- 4 0. ings.Namely,whenrecognitionwasnotdiagnosticorun- No cue Congruent cue Incongruent cue available, the average confidence level was 60.42% (SD = 6.4). Confidence level increased to 66.3% (SD = 9.4) 3 Results whenrecognitioninformationdifferentiatedbetweenop- tions. A within-subject repeated measures ANOVA re- vealed that this difference was significant [F (1, 23) = 3.1 Recognitionandcuevalidityestimates 19.178,p<0.0001]. On average, participants recognized approximately 9 Taken together, these results support the first part of cities (SD = 1.5; range 5–12). A one-way analysis of theRHhypothesesaccordingtowhichdiagnosticrecog- variance (ANOVA) revealed that the order of this task nition information leads to high choice proportions that (i.e., beforeoraftertheexperimentaltask)hadnoeffect fit the recognition cue, short RT, and a high confidence onrecognitionranks[F(1,23)=0.259,p=0.616]. This level (see Table 1). The results also support the validity resultsuggeststhatparticipantswereabletodifferentiate ofourmeasuresinthistask. theirexperience-basedrecognitionfrommerefamiliarity with the city names that was induced by repeated expo- 3.2 Choiceproportionanalysis suretothecitynamesintheexperiment. The validity of the recognition cue across all critical The results of the choice proportion analysis are pre- trials in which only one city was recognized was 77% sented in Figure 1. As mentioned above, participants’ (SD=12.6%),andsignificantlyabovechancelevel[t(23) choices followed recognition in, on average, 83.3% of = 10.724, p < 0.0001 in a one-sample t-test]. This anal- thecasesinwhichnoinformationotherthanrecognition ysis confirms the ecological rationality of using the RH was presented. This proportion increased to 95.1% (SD in the current study, establishing the task as a fair test = 6.8%) when a cue congruent with the recognition in- forexaminingtheprocessesunderlyingtheprocessingof formationwasintroducedanddecreasedto50.1%(SD= recognition information. Importantly, the validity of the 26%)whenanincongruentcuewaspresented. Awithin- recognitioncuewasfarabovethe60%validityofthead- subjectrepeatedmeasuresANOVAwasconductedtotest ditional cognitive cue that was provided in some of the the effect of cue type (i.e., no cue, congruent cue, and trials(i.e.,informationfromthegeographystudent). incongruent cue) as the independent variable on choices Next,weexaminedwhethertheuseofrecognitionin- which followed recognition as the dependent variable. formationaffectedchoicebehavior. Todoso, wecalcu- This analysis revealed a significant main effect of cue latedtheproportionofchoicesthatfollowedtherecogni- type [(F (2, 69) = 59.26, p < 0.0001). Post-hoc paired- tioncue,andcomparedittochancelevel.Aone-sampled sample analyses demonstrated that the no-cue choices t-test showed that choices were sensitive to the recogni- differedsignificantlyfromboththecongruentandthein- tion cue in the majority of cases in which only recog- congruent cue choices [t(23) = -4.919, p < 0.0001, and nition information was available [average = 83.3%; SD t(23)=7.265,p<0.0001,respectively]. = 13%; t(23) = 12.889, p < 0.0001]. In addition, we examined whether response times and confidence level 3SincenodifferencewasfoundbetweentheRTsforeitherrecog- were affected by recognition information. Although the nizedornone-recognizedcitypairs,theseresultswerecollapsed. JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 292 on only one piece of information in a noncompensatory Figure2:A-Meanresponsetimeasafunctionofthetype manner. Instead, theresultssuggestthatparticipantsare of information available besides recognition. B - Mean sensitive to all information, and base their decisions on confidencelevelasafunctionofthetypeofinformation itsintegration. available besides recognition. Error bars represent stan- Partially in contrast to previous findings (e.g., Gold- darderrors. stein & Gigerenzer, 2002; Pachur et al., 2008), it seems 5 9. A that recognition information was completely overridden by other information, even when the additional cue was of low validity and was obtained from a relatively un- 0 known source. In fact, when a cue — with a validity as c) 9. lowas.60—thatwasincongruentwithrecognitionwas e s e ( presented,iteliminatedtheeffectoftherecognitioncue, m so that the proportion of choices following recognition se ti 8.5 droppedtochancelevel. Thus,inlinewithmostpriorre- n po search,thecurrentresultssuggestthatrecognitioninfor- s Re mation is only one of several cues that are integrated in 0 acompensatorymanner. Moreimportantly, ourfindings 8. suggest that real experience-based recognition informa- tionisnotsignificantlystrongerthanevenalowvalidity cognitivecue. Finally,assuggestedbypreviousresearch(e.g.,Pachur No cue Congruent cue Incongruent cue et al., 2008), individual differences might exists in peo- 2 ple’suseofrecognitioninformationandadditionalcues. 7 Toexaminethisalternativespeculationwecalculated,for B each individual, the correlation coefficient between pro- 0 7 portionsofchoicesthatfollowedrecognitionandcondi- tion (i.e., congruent cue, no cue, and incongruent cue). %) el ( 68 The median correlation coefficient between choice and v conditionacrossparticipantswas–0.95(t(23)=–16.406, e e l p < 0.05). Similarly, for 95.6% of the participants the enc 66 correlationturnedouttobehighlysignificant(p<0.05). d nfi Thispatternofindividualresultssupportstheaggregated o C 4 results and suggests that choice proportions were high- 6 estwhenacongruentcuewaspresentedandlowestwhen an incongruent cue was presented even at the individual 2 6 level. Thereby,individualdifferencescannoteasilyserve as alternative explanation for our results, as there seems No cue Congruent cue Incongruent cue tobeonlysmallindividualdifferencesintheuseofrecog- nitioninformationandadditionalcues. These results replicate previous findings (e.g., Bröder 3.3 Responsetimeanalysis & Eichler, 2006; Oppenheimer, 2003; Richter & Späth, 2006) and support the qualitative predictions concern- TheanalysisofresponsetimesispresentedinFigure2A. ing changes in choice proportion derived from the PCS Overall,themeanresponsetimeforcasesinwhichonly model. Theresultsconflictwithpredictionsderivedfrom one city was recognized was 8.48 seconds (SD = 0.88) theRH.Specifically, theresultssuggestthatrecognition whennootherinformationwaspresented.Responsetime information is integrated with additional information in decreased to 8.11 seconds (SD = 0.84) when a congru- a compensatory manner. The claim that recognition in- ent cue was introduced and increased to 9.14 seconds formation is used as the only information in a noncom- (SD = 1.32) when an incongruent cue was presented. pensatorymannerhastoberejected. Moreover,thispat- A within-subject repeated measures ANOVA was con- tern of results does not even support a weaker interpre- ducted to test the effect of cue type (i.e., no cue, con- tation of the RH (Brandstätter, Gigerenzer, & Hertwig, gruent cue, and incongruent cue) on response time as 2008),whichassumesthat,whiledecisionmakersevalu- the dependent variable. This analysis revealed a signif- atealltheavailableinformation,thefinalchoiceisbased icant main effect of cue type [(F (2, 69) = 17.472, p < JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 293 0.0001). Post-hoc paired-sample analyses demonstrated 9.4%)whennootherinformationwaspresented. Confi- thatresponsetimeswithoutanycuedifferedsignificantly dencelevelincreasedto69.4%(SD=10.1%)whenacon- fromresponsetimesfollowingboththecongruentandthe gruentcuewasintroducedanddecreasedto64.3%(SD= incongruent cue [t(23) = 2.692, p = 0.01, and t(23) = – 7.9%)whenanincongruentcuewaspresented.Awithin- 3.136,p=0.005,respectively]. subjectrepeatedmeasuresANOVAwasconductedtotest Thus,similartothechoiceproportionresults,decision the effect of cue type (i.e., no cue, congruent cue, and time results support the predictions of the PCS model, incongruent cue) on confidence level as the dependent replicating findings by Hilbig and Pohl (2009). In line variable. Thisanalysisrevealedasignificantmaineffect with the predictions of the PCS model, our findings of cue type [(F (2, 69) = 25.361, p < 0.0001). Post-hoc demonstrate that when cue predictions are incongruent paired-sampleanalysesdemonstratedthattheconfidence (i.e., when recognition information points at one choice levelintheno-cuechoicesdifferedsignificantlyfromthe andthecuepointsattheoppositechoice),responsetime confidencelevelofboththecongruentandtheincongru- isincreased. Furthermore,wefoundsupportforthePCS entcueconditions[t(23)=-4.791,p<0.0001,andt(23) hypothesis that decision times are decreased if the addi- =2.73,p=0.01,respectively]. tional cognitive cue points in the same direction as the Again, the results support the predictions of the PCS recognitioncue.Thisfindingisimportantinshowingthat model but not the predictions derived from the RH. In more information (i.e., the recognition cue + the cogni- particular,ourfindingssuggestthatavailableinformation tive cue) is processed faster than less information (i.e., isintegratedandthatitaffectsconfidencelevel.Theanal- recognitioncuealone). Theseeffectsspeakagainstserial ysis of confidence level shows that rather than relying modelsofinformationintegration(e.g.,Payne,Bettman, solely on recognition information, decision makers con- &Johnson,1988;seealsoGlöckner&Betsch,2010). siderrecognitiontobejustoneofothercuesthatareinte- Moreover, ifrecognitioninformationwereconsidered gratedbyacompensatoryprocessonthewaytoreaching in a noncompensatory manner, when no additional cue adecision. was presented, response times should have been equal At the individual level, the median correlation coeffi- to the other two conditions (either congruent or incon- cientbetweenconditionandconfidencelevelacrosspar- gruent cues) or even shorter. Shorter response times for ticipants was –0.89 (t(23) = –13.599, p < 0.0001), sug- recognition alone could support a weaker interpretation gesting that confidence level was highest for congruent ofthenoncompensatoryviewpoint,inwhichallavailable trialsandlowestforincongruenttrials.Similarly,for92% informationisscannedbutthefinaldecisionisbasedon oftheparticipantsthiscorrelationturnedouttobehighly thebestcuealone. Insuchcases,decisionmakerswould significant(p<0.0001). Thus, thisresultaddsuptothe havelessinformationtoevaluateandcouldthusbefaster. choiceproportionsandresponsetimesfindings, andfur- However,theresultsprovidenosupportforthisweakin- thersupporttheclaimthatthecurrentresultsdonotstem terpretationofthenoncompensatorymodels. Rather,the from individual differences in the use of recognition in- patternofresultsareinlinewithPCSmodels,suggesting formation. that more processing time is required in cases in which itishardertofindaconsistentmentalrepresentationdue 3.5 Physiologicalarousal tohighconflict(i.e.,lowconsistency)betweenpiecesof information. 3.5.1 Emotionalresponsetorecognizedstimuli Finally,asforchoiceproportions,wecalculated,atthe individual level, the correlation coefficient between re- To evaluate whether recognition information creates an sponse time and condition. The median correlation co- affective signal, we compared physiological reaction, efficientbetweenresponsetimesandconditionwas0.93 measuredforthetaskinwhichparticipantsratedwhether (t(23)=9.235,p<0.05),suggestingthatattheindividual theyrecognizedthecityornot(forsinglecities). Hence, level response times were highest for incongruent trials wemeasuredphysiologicalreactioninresponsetorecog- and shortest for congruent trials. Similarly, for 86% of nized city names versus unrecognized city names. Note theparticipants,thiscorrelationwashighlysignificant(p thatlowPATmeasuresindicatehigharousal. Ascanbe < 0.05). Thus, as in choice proportions, this result ex- seen in Figure 3A, there was a slight tendency that rec- cludesanindividualdifferencesaccountofourdata. ognized cities led to higher arousal, particularly 2 to 4 secondsafterstimulusonset.However,thistendencywas 3.4 Confidencelevelanalysis notstatisticallysignificant. From2secondspriortocity name presentation to 7 seconds after presentation, there Theresultsoftheconfidencelevelanalysisarepresented was no significant difference between PAT responses to inFigure2B.Overall,themeanconfidencelevelforcases the different types of stimuli. Namely, the average al- inwhichonlyonecitywasrecognizedwas66.3%(SD= phaacrossallwithin-subject(recognizedversusunrecog- JudgmentandDecisionMaking,Vol.5,No.4,July2010 Physiologicalarousalinprocessingrecognitioninformation 294 0.91 in this analysis, making it unlikely (i.e., beta-error Figure 3: A — Average PAT scores in response to rec- = 0.09) that an existing effect was merely not detected ognized versus unrecognized city names. B — Average (Faul,Erdfelder,Lang,&Buchner,2007). PAT scores as a function of information type. Low PAT Althoughnullhypothesisresultsshouldbeinterpreted scoresindicatehigharousal. cautiously, our findings imply that in the current task recognitioninformationdoesnotseemtocreateanaffec- 8 A tivesignalfavoringtherecognizedcity,oritmaycreatea 7. veryweaksignalwhichishardtodetect.Thisconclusion shouldbefurthervalidatedinfutureresearch. Unrecognized 6 7. Recognized s) olt 3.5.2 Physiological response to additional informa- v AT ( 7.4 tion P To examine physiological response to additional cue in- 2 formation, we compared PAT scores in response to con- 7. gruent versus incongruent trials.5 As mentioned above, PAT scores represent vasoconstriction, and low scores 7.0 thereforeindicateincreasedarousal. Ascanbeseenfrom Figure 3B, arousal was higher in incongruent trials than −2 −1 0 1 2 3 4 5 6 7 in congruent trials (i.e., lower PAT scores). A within- Time (relative to city name presentation subjectrepeatedmeasuresANOVArevealedthatthisdif- in the recognition task) ference was significant at one second after the presenta- tionofinformation[F(1,23)=4.09,p=0.05]. 2 B Thus, in line with the PCS model prediction, arousal 8. Incongruent increased(i.e.,lowerPATscoreswereobtained)withde- 0 Congruent creasing consistency between the available information. 8. Thearousaldatafurthersuggestthatadditionalinforma- tion is integrated rather than neglected, as would have olts) 7.8 been predicted by the RH. Specifically, the increased v PAT ( 7.6 acroonussisatlenincyr)e,sbpuotnnseottoincroenspfloicntsinegtoinrefocromgnaittiioonn(iin.efo.,rmloaw- tion alone, suggests that information is integrated in a 4 7. compensatorymannerandthatrecognitioncuesareused asoneofseveral(cognitive)cues. 2 7. 4 Discussion −2 −1 0 1 2 3 4 5 6 7 Time (relative to pair of cities and additional cue presentation in the city−size task) Thecurrentstudyaimedtoexaminethenatureofrecog- nition information in human inference tasks, as well as the integration of such information with additional cog- nized)repeatedmeasuresANOVAwas0.51(SD=0.25, nitive cues. In particular, we examined two main re- range0.18–0.99).4 Wecannot,however,ruleoutthepos- search questions. First, we tested inferences regarding sibility that the results might be partially due to the rel- ativelysmallsizeofoursample. Still, apost-hocpower 5ThePATscoreishighlysensitivetocognitiveload(Iani,Gopher, analysis(for10citieswithineachgroup,withanassumed Grunwald,&Lavie,2007;Iani,Gopher,&Lavie,2004),whichisas- correlationofr=0.3betweenobservationsandassuming sumedtoincreaseasthetaskprogresses. Tocontrolforthiseffect,we asmalleffectoff=0.25)indicatedanexcellentpowerof matchedforeachparticipantthenumberoftrialsineachcondition(con- gruentandincongruent),sothateachconditionhadthesamenumberof trialsastheconditionwiththeminimalnumberoftrials(Mean=17 4TheresultsoftherepeatedmeasureANOVAforthedifferencebe- trials).Itshouldbenoted,however,thatwhenweregressedoutthetime tweenthePATscoreforrecognizedvs.unrecognizedcities2to5sec- trend(i.e.,totaltrials)fromthePATscoresandusedregressionanalysis ondsafterstimuluspresentationswereF2sec(1,23)=1.941,p=0.18; topredictthePATscorefromthetypeofcue(congruentversusincon- F3sec (1, 23)=0.933, p=0.34; F4sec (1, 23)=0.727, p=0.40; and gruent)wefoundsimilarpatternofresults,althoughtheeffectofcue F5sec(1,23)=0.741,p=0.40. (Wedidnotcalculatethetestat1sec typewasonlymarginallysignificant(i.e.,thep-valueofthemodelwas becausetherewasnodifference,seeFigure3A). 0.055;one-tailedtest).

Description:
ing confidence ratings, response times, and choice proportions further demonstrated that recognition information and other cognitive cues are integrated in a compensatory manner. Keywords: affect, recognition heuristic, physiological arousal, parallel constraint satisfaction, noncompensatory models
See more

The list of books you might like

Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.
DMCA & Copyright: Dear all, most of the website is community built, users are uploading hundred of books everyday, which makes really hard for us to identify copyrighted material, please contact us if you want any material removed.
Copyright @ PDFdrive.to, 2022 | We love our users