634Reasoning and Problem Solving
instancesofrainorifpeopleemployabiasedsearch,inwhich
theyignorealloflastweek’ssunnydaysandfocusonlyon
therainydaysinassessingtheprobabilityofraintomorrow
(seethechapterbyNickerson&Pewforafullerdiscussionof
heuristics).TverskyandKahneman(1974,1986)considered
thatpeople’srelianceonheuristicsunderminedtheviewof
peopleasrationalandintuitivestatisticians.Otherinvestiga-
torsdisagree.
Fast and Frugal Heuristics
Gigerenzer,Todd,andtheircolleaguesfromtheABCresearch
group(1999)havesuggestedthatpeopleemployfastandfru-
galheuristicsthattakeaminimumamountoftime,knowl-
edge,andcomputationtoimplement,andyieldoutcomesthat
areasaccurateasoutcomesderivedfromnormativestatistical
strategies.Gigerenzeretal.(1999)haveproposedthatpeople
usethesesimpleheuristicstogenerateinferencesineveryday
environments.Onesuchheuristicexploitstheefficiencyof
recognitiontodrawinferencesaboutunknownaspectsofthe
environment.Inadescriptionoftherecognitionheuristic,
Gigerenzeretal.statedthatintasksinwhichonemustchoose
betweentwoalternativesandonlyoneisrecognized,therec-
ognizedalternativeischosen.Asthisstatementsuggests,the
recognitionheuristiccanbeappliedonlywhenonealternative
islessrecognizablethantheotheralternative.
Inaseriesofexperiments,Gigerenzeretal.(1999)showed
thatpeopleusetherecognitionheuristicwhenreasoningabout
everydaytopics.Forexample,inoneexperiment,21partici-
pantswereshownpairsofAmericancitiesplusadditionalin-
formationabouteachofthecitiesandaskedtochoosethe
largercityofeachpair.Theresultsshowedthatparticipants’
choicesoflargecitiestendedtomatchthosecitiestheyhadse-
lectedinapreviousstudyasbeingmorerecognizable.The
recognitionheuristiccanoftenleadtoaccurateinferencesbe-
causeobjectsorplacesthatscoreveryhigh(orverylow)ona
particularcriterionarenormallymadesalientinourenviron-
ment;theiratypicalcharacteristicsmakethemstandout.
Therecognitionheuristicalsoyieldsaccurateinferencesin
businesssituationssuchasthosethatinvolvestockmarket
transactions.Inonestudy,480participantsweregroupedinto
oneoffourcategoriesofstockmarketexpertise—American
laypeople,Americanexperts,Germanlaypeople,andGerman
experts—andaskedtocompleteacompanyrecognitiontask
ofAmericanandGermancompanies(Gigerenzeretal.,
1999).Participantsthenmonitoredtheprogressoftwoin-
vestmentportfolios,oneconsistingofcompaniestheyrecog-
nizedhighlyintheUnitedStatesandtheotherconsistingof
companiestheyrecognizedhighlyinGermany.Participants
analyzedtheperformanceoftheinvestmentportfoliosfora
periodof6months.Resultsshowedthattherecognition
knowledgeoflaypeopleturnedouttobeonlyslightlyless
profitablethantherecognitionknowledgeofexperts.Forin-
stance,theinvestmentportfolioofGermanstocksbasedon
therecognitionoftheGermanexpertsgained57%duringthe
study;however,Germanstocksbasedontherecognitionof
theGermanlaypeoplegained47%duringthesameperiod—
only10%lessthanthegainsmadebymeansofexpertadvice!
TheinvestmentportfoliosofU.S.stocksbasedontherecog-
nitionofAmericanlaypeopleandexpertsdidnotmakesuch
dramaticgains(13vs.16%,respectively).However,inall
cases,portfoliosconsistingofrecognizedstocksyieldedaver-
agereturnsthatwere3timesashighasthereturnsfrom
portfoliosconsistingofunrecognizedstocks.Thesefindings
indicatethatwhenoneisinvesting,asimpleheuristicmight
beaworthwhilestrategy.Probability Heuristic ModelAnotherheuristicapproachtoreasoningisChaterand
Oaksford’s(1999)probabilityheuristicmodel(PHmodel)
ofsyllogisticreasoning(seealsoOaksford&Chater,1994).
AccordingtoChaterandOaksford,simpleheuristicscan
accountformanyofthefindingsinsyllogisticreasoning
studieswithouttheneedtopositcomplicatedsearch
processes.InthePHmodel,quantifiedstatementssuchas
AllbirdsaresmallorMostapplesareredareorderedbased
ontheirinformationalvalue.Usingconvexregionsofa
similarityspacetomodelinformativeness,Chaterand
Oaksfordshowedmathematicallythatdifferentquantified
statementsvaryinhowmuchspacetheyoccupyinthesim-
ilarityspace.Categoriessuchasallandmostinquantified
statementsoccupyasmallproportionofthesimilarityspace
andoverlapgreatly,andarethusconsideredmoreinforma-
tivethanthosequantifiedstatementswhosecategories
occupyalargerproportionofthesimilarityspaceanddo
notoverlapgreatly(seetheirAppendixA,p.242).Inother
words,quantifiedstatementsconsideredtobehighininfor-
mationalvaluearethose“thatsurpriseusthemostifthey
turnouttobetrue”(Chater&Oaksford,1999,p.197)
becauseweperceivethemasunlikely.InChaterand
Oaksford’s(1999)computationalanalysis,quantifiersare
orderedasfollows:All>Most>Few>Some...are>No...are>>
Some...are notThus, statements containing the quantifier all,such as All
people are tall,are considered more informative than state-
ments containing the quantifier most,such as Most people are
tall.where>stands for more informative than.