Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Attitudes as Predictors of Behavior Versus Behavior Intentions: A Convergence of Research

Traditions
Author(s): Stan L. Albrecht and Kerry E. Carpenter
Source: Sociometry, Vol. 39, No. 1 (Mar., 1976), pp. 1-10
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2786586 .
Accessed: 25/06/2014 10:39

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
Sociometry.

http://www.jstor.org

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
Sociometry
4976, Vol. 39, No. 1,,1-10

Attitudesas PredictorsofBehavior
VersusBehaviorIntentions:A
ConvergenceofResearchTraditions
STAN L. ALBRECHT
YoungUniversity
Brigham
KERRY E. CARPENTER
StateUniversity
Washington

Two researchmodelsfor the studyof attitude-behavior relationshipsare reviewed and their


and differences
similarities discussed.The first,attributed primarilyto sociologistM. L.
DeFleur and his associates,is characterized as a "contingent consistency" approachto
attitude-behaviorstudy.Thesecond,whichis largely theworkofpsychologist M. Fishbeinand
his associates,redefinessome earlierconceptsand emphasizesthe variableof "behavior
intentions" (BI). In an empiricalevaluationof the two approaches, measures are takenof
attitude,behaviorintentions, normative beliefsand behavior.Themultiple using
coefficients
attitudeand normative beliefto predictbehaviorare quitehigh,thoughthetwoindependent
are highlyinterrelated.
variables In one instancetheBI-Brelationship is strongerthantheA-B
relationship,thoughnot in another.Possiblereasonsfor thisdifference are discussed.It is
concludedthattheFishbein-type model,ifmodified, neednotbe confined to narrowly defined
norto behavior
attitudes, intentionsratherthanbehavior. thata combination
It is suggested of
thetwoapproaches might infuture
provefruitful attituderesearch.

Whilemuch of the work in the area of contention,some observationsabout the


attitude-behaviorresearchhas dealt with xesearch from which the Wicker and
problemsofattitudedefinition,measurement, Deutscherconclusionswere largelydrawn
and change,one of themostwidelydebated becomecritical.In the firstplace,mostifnot
problemsof recent years is that of the all of the earlystudiesdealt withbivariate
relationshipbetween attitudes and overt relationships. Thatis, attitudewasusedas the
behaviortowardtheobjectoftheseattitudes. sole predictorof behavior.It is somewhat
For several decades, beginningwith the curious that it took at least two decades
pioneeringeffortof LaPiere (1934), social before sociologistsbegan to recognizethe
havedebatedtheefficacy
scientists ofattitude importance ofgoingbeyondbivariate relation-
as a predictorof overt behavior.Careful ships and to take into accountsociological
reviewsof the researchaccumulatedduring (social situational)factorsin understanding
this period led Wicker(1969) to conclude theattitude-behavior relationship.
Theirtrain-
that,at best,attitudeaccountsforaboutten ing as sociologistsshouldhave warnedthem
percent of the variabilityin predicting much earlier that to rely strictlyon
behavior.Deutscher(1966, 1973) has gone "predispositional" factorsin understanding
even furtherin arguingthat there is no behaviorwouldbe problematic.
theoreticalreason to expect congruence By the late fifties, however,DeFleurand
betweenwordsand deeds,and,in fact,every Westierecognizedthe importanceof social
reasonto expectdiscrepancies. situationalvariables in understanding the
However,examinationof the attitude-be-attitude-behavior
problem.2In theireffort to
haviorresearchwhichhas accumulatedsince
LaPiere'sarticle,showsthatsomeimportant
there have been significantadvances, though
steps have been taken.' In supportof this probablynot chronologically
and certainlyin a
sometimes andunsurefashion.
hesitating
IThis statement
is not madewithouta certain 2Whilesomeearlierwriters 1952)
(e.g.,Minard,
degreeof reservation. (1969) quite acknowledgethe problem,this recognition
WhileDeutscher was
seriouslyquestionstheamountof progress
thathas seldom translatedfrom their rhetoricto their
reallybeenmadesinceLaPiere,we tendto feelthat research.
1

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
2 SOCIOMETRY
explain the behaviorof subjects who ex- DeFleurhas devotednearlytwodecadesof
hibitedattitude-action inconsistency,DeFleur work to the problem of attitude-behavior
andWestiestate: relationships(DeFleur and Westie, 1958,
The lack of a straight-line relationship 1963; Warnerand DeFleur,1969; Albrecht et
betweenverbalattitudeand overtaction al., 1972; Acock and DeFleur, 1972).3 From
behaviormorelikelymay be explainedin this researchhas evolved a "contingent
termsofsomesortofsocialinvolvement of consistency"approach to the study of
the subject in a systemof social con- attitudesandbehavior.Thisapproachsuggests
straints,preventing himfromacting(overt- that various social constraints(such as
ly) in the directionof his convictions, or perceived group norms and visibilityof
otherwise"legitimizing" certainbehavioral behavior)pose contingent conditionswhich
patterns.These channelizing influenceson
behaviorhave receivedtheoreticalatten- may act to modify the strengthof the
tion in terms of such concepts as attitude-behavior relationship (Warnerand
"referencegroups,""otherdirectedness," DeFleur, 1969). Varioustypesof intervening
and "significant others"(1958:672). variables mayaffecttheobservedrelationships
A numberof researchers, influencedby and the principalresearchtaskbecomesone
this insight,have operationalized these and of identifying and measuringthe effectof
variousother"situational constraints" foran these variables. More recently,Acock and
improvedunderstanding of attitude-behavior DeFleur have introduced a "configurational
relationships. This has led, in turn,to the approach"to thebasiccontingent consistency
development of at least two majorresearch model. The configurational approach assumes
that " .. .. when attitude is combined with
traditionsin sociologyand socialpsychology
for the understandingof attitudes and social influencevariablesin configurations
behavior.The firstof these empiricalap- involvinginteractioneffectsbetween such
variables,(theoretically) thepredictive power
proacheshas been developedby and has
is greaterthan whenthesevariablesoperate
evolvedout of the workof sociologistM. L.
DeFleur and associates. The second is separately" (1972:716).
The basicmodelforresearchsuggested by
attributedprimarilyto psychologistM. the
contingent consistency approachinvolves
Fishbeinand his associates.The approaches
combining attitudewithsetsofindividual and
have severalthingsin common.For example,
social constraint variablesforthe prediction
bothhaveclearlymovedbeyondthestudyof of
behavioraloutcomes. Attitudeis then
bivariaterelationships and recognizethatthe viewedas
but one ofthecomponents ofwhat
problemat handis muchmorecomplex.Both will
usuallybe a rathercomplexequationfor
traditions also recognize thecontinued impor- determining action.The configurational idea
tance of attitude in their multivariate
suggeststhattherelationships are not simply
relationship. Thatis, ratherthanrejecting the additivebut also
involveinteractions between
use of the attitudeconceptaltogether, they the
variables.
have continuedto workwithit, refining it
One of the basic problemswith this
when necessary,and combiningit with
setsof situational approach is that of establishing parsimony.
different and othervariables
untilreachinga muchhigherlevelofbehavior Thatis, how does one determine whichsocial
predictionthanhas been approachedin past constraintvariablesare most crucial and
research.In sum,bothDeFleurend Fishbein requireinclusionin theresearch process?One
recognizethatbehaviorin anygiveninstance of the consequences of widespread acceptance
is likely to be the result of multiple of thecontingent consistency notionhasbeen
determinants, attitudes. a rapid multiplication of "intervening vari-
including
The two approaches differin several ables" that may act as constraints on the
criticalrespects.It is thesedifferencesthatare attitude-behavior relationship.4 Parsimony de-
of primaryconcernhere. We will beginby
brieflydescribing theprincipalcharacteristics 3In additionto DeFleur's own studies,the work
of the two approachesin orderto pointout of Linn (1965), Fendrich(1967), DeFriese and Ford
theirdifferences, andwillthenpresentdatain (1968), as well as others,reflectsthe influenceof his
an attemptto pointoutareasin whichgreater effort. 4This problem has also been suggested by
convergence maybecomepossible. Schwartz and Tessler who note that the principle

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
ATTITUDES,BEHAVIORAND BEHAVIORINTENTIONS 3
mands that these be limited to some where B = overt behavior, BI = behavior
manageableset. Acock and DeFleur (1972) intention, Aact = attitudetowardperforming
suggestthat perceivednormsof familyand the specificact in question,NB = normative
peers,combinedwithattitude,offersa high beliefs,and MC = motivation to complywith
level of prediction of marijuana-relatedthe normativebeliefs. Fishbeinnotes that
behavior in a laboratory context. while additionalfactorsmay contributeto
This emphasison attitude,combinedwith predictionof behaviorintentions, theydo so
some operationalmeasureof social norms, only insofaras they influenceone of the
bringsus to a discussionof Fishbein'swork. specificfactors in thebasicmodel.
Fishbeinand his associates(Fishbein,1963; In termsof supportivedata, it has been
1967; 1973; Azjenand Fishbein,1969; 1970; foundconsistently thatmultiplecorrelations
1972; 1973; Ajzen, 1971; Darroch,1971; of Aact and NB(MC) on BI are veryhigh.
Hornik,1970; Carlson,1968) havedeveloped Azjen and Fishbein(1973) reportthat the
a theoryof attitude-behavior relationshipsaveragemultiplecorrelationin ten different
thatattemptsto integrate a set of predictive studiesis about .81 (R2 = about .66). Thisis
variablesinto a singleconceptualframework.an impressivefinding,especiallywhen one
The theorydeals with threebasic variables considersWicker's(1969) conclusionfromhis
that are viewed as being antecedentto extensivereviewof the literaturethat the
behaviorintentions (not behavior perse). The amountof variationin behaviorthatcan be
three basic antecedentvariables are (1) explained by attitude averagesabout 10
attitude toward performingthe act or percent.However,as will be noted below,
behaviorin question (Aact), (2) normative Fishbein is dealing with a very different
beliefsabout thatbehavioralact (NB)5, and variablethanis foundin mostof thestudies
(3) motivation to complywiththosebeliefs reviewedby Wicker.Thisbasicmodeloffers a
(MC). The two specific determinants of high degreeof parsimonyand is similarto
behaviorintentionsincludean "attitudinal" DeFleur's emphasison attitudecombined
factorand a "normative"factor(the latter withperceived peerand family norms.
being a product of normativebeliefsand SeveralthingsaboutFishbein'sformulation
individualmotivationto complywiththose requirespecialcomment, in terms
particularly
beliefs). of their relationshipto models that have
In symbolform,the centralequationthat governed the work of a majority of
guidestheworkof Fishbeinandhiscolleagues sociologists who have dealtwiththeattitude-
is written as follows: behaviorproblem:
B t BI = (Aact)WO+ [NB(MC)] w16 1) In the firstplace, Fishbeinis directly
concernedwithpredicting behaviorintentions
and not the traditionalkinds of overt
problemwiththe "othervariables"explanations of behaviorswith whichmostsociologists have
attitude-action discrepancy"is thatthe numberof
casualvariablespotentially worthy of consideration dealt.He notes thatin manycases BI and B
isvirtually infinite" (1972:325). will be stronglycorrelatedand that, as a
5In themodel,normative beliefs(NB) refers to consequence, theequationwillalso providea
"the actor's belief about the likelihoodthat strongpredictionof behavior.However,such
members of a givenreference groupexpecthimto a strongrelationship need not alwaysbe the
performthe behaviorin question" (Azjen and
Fishbein,1973). It is recognizedthat different case. The centralpoint forthepresentpaperis
reference groupsmaybecomerelevantin different thatthe thrustof Fishbein'sworkis toward
situations and thatin someinstancesthe expecta- predictingbehaviorintentionsand not be-
tions of more than one reference groupmay be havior.
salientto a particular behavioraldecision.In the
lattercases,the predictive equationis expandedto Accordingly, severalcriticalproblemsare
accountforadditional referent
expectations. left unresolved. For example, it seems
6 Thoughthemodelcallsformultiplying theNB reasonableto arguethatthe centralconcern
factorby motivation to comply(MC), Fishbeinand of the social sciencesis stillwithpredicting
hiscolleaguesadmitgreatdifficultyinoperationaliz-
ingthislattervariable.
As a consequence,
itis almost
alwaysomittedfromthe finalanalysis.Theynote
thatthisvariablecan be viewedin twoways:1) "it referent'sparticulardemands," or 2) "it is possible
can be seen as the person'smotivationto comply to view it as specificto the givenexpectationof a
with a givenreferencegroup,regardlessof the referencegroup" (Azjen and Fishbein 1973:43).

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
4 SOCIOMETRY
and understanding human behavior. The (and frequentlytrivial) behaviorsbut be
originalpopularityof the attitudeconcept unableto generalizefromthoseto otheracts
resulted,in largepart,fromthebeliefthatit in othertypesof situations.
would help to predict and understand 3) Sociologistsstudyingattitude-behavior
behaviordirectedtowardthe object of the relationships have traditionally used Likert-
attitude.It seemsfarmoreimportant to know type scales to measureattitudetowardthe
one's actual behaviorin relationship to such object. Fishbein,on the otherhand,uses the
issues as marijuanause and legalizationor semanticdifferential formeasuring attitudes.
fertilitypatterns,ratherthan simplyone's For example,one study(Ajzen and Fishbein,
statedbehavioralintentions. The latterleaves 1972) dealtwitha seriesof hypothetical acts
us with the same criticalproblemthat has (such as buyinga plot of land) to which
plaguedattituderesearchers in recentyears: subjectswere asked to respond.To measure
whatarethefactorsthatintervene betweenBI Aact, buyingthe plot of land was ratedon
andB to strength or weakenthatrelationship? four evaluativesemanticdifferential scales
In otherwords,whiletheBI equationsgiveus with the followingadjective alternatives:
impressivecoefficients,we still face the foolish-wise,good-bad, harmful-beneficial,
criticalB - BI problem;we muststillconcern rewarding-punishing. The sum across these
ourselveswiththeproblemsof whatmensay fourscaleswasthenusedas theindexofAact.
they will do and what they otherwisedo Again, the specificityof the attitude
(Deutscher,1973). As a consequencethenext measureis critical.The rangeof behavioral
generationof social psychologistsmay be actsin whichone canengagein relationship to
promptedto discoverthosefactorsthatmake an attitudeobject (for example,blacks) is
theB-BI relationship a "contingent" one,just almost infinite.It seems crucial that we
as we have been doing with the A-B developmeasuresthatcoverat leasta partof
relationship.7 thathypothetical range.Predictionof a very
2) The firstindependentvariablein the specific act may be less importantthan
Fishbeinequationis attitudetowardperform- knowinga generalrangeofresponseprobabili-
ing a specific act under specificcircum- ties.
stances-not attitudetowardan object. In The Acock-DeFleur paperis similarenough
otherwords,Fishbeinmeasuresnot attitudes to Fishbein'sformulation thatit suggeststhe
towardblacksbutattitudetowardengaging in model may have far broaderapplicability.
a specificact in relationship to blacks.It has That is, perhapsone can deal with a more
longbeenrecognized thatone of theproblems generaland traditional attitudemeasurement
in past attitude-behavior researchis thatthe procedureas well as withactualbehavior(as
researcher hasmeasureda generalattitudeand opposedto behaviorintention)andstillcome
thengrievedthatit is a poor predictorof a up witha highlevelofbehaviorpredictability.
specificbehaviortowardthe attitudeobject. We test this hypothesisby employing
Fishbein does not have this problem;the Fishbein'sbasic modelto re-examine thedata
attitudetowardthe act deals withthe very used in therecentAcock-DeFleur article.An
specificbehaviorthat is to be predicted. additionaltest includesa second behavioral
However,both the BI and the Aact have a measure obtainedin a follow-upinterview
tendencyto become so specificthat the with a sample of subjectsincludedin the
abilityto generalizefromany of the findings Acock-DeFleur report.8
is severelycurtailed.In other words, in
improving thesizeof ourcoefficients, we may ResearchProcedures
comeup withexcellentpredictions of specific Researchproceduresemployedin thisset
coveredby Acockand
of studiesare generally
7Ajzenand Fishbein(1973), forexample,note DeFleur (1972). Several hundred under-
thattheshorter thetimespanbetweenB andBI, the graduatestudentsat a major northwestern
stronger shouldbe the relationship.
The samehas university completeda ratherlengthyques-
been said about the A-B relationship
(see Ehrlich,
1969). In fact,probablymostof the factorsthat
havebeensuggested as variables
intervening
between 8It shouldbe notedthatthistestwillbasically
attitudeand behaviorare equallyreasonablewhen involvean analysiswhichutilizesthe independent
lookingat the behavior-behavior relation- variablesfromthe Fishbeinmodel to predictthe
intention
ship. dependent model.
variableintheDeFleur-type

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
ATTITUDES,BEHAVIORAND BEHAVIORINTENTIONS 5
tionnaireat the beginningof fall term. legalization by checking a privateballotwhich
Includedin the questionnairewerea seriesof wasthenplacedin a "secretballotbox."
attitudescales, includingone designedto In additionto the above phases of the
measure attitudes toward marijuana.The researchthat are coveredby the Acock-De-
questionnairealso obtainedextensiveback- Fleurreport,an additionalstepwas addedto
grounddata on the subjectsand included the researchprocedurethat is of special
personalitymeasuresdesignedto get at such interesthere. A few weeks followingthe
thingsas need forsocialapprovaland degree laboratorystudy,a randomsampleof 59 of
of inneror otherdirectedness. Data werealso the subjectswas contactedfor a follow-up
obtainedon important reference groupsand interview. Duringtheinterview, subjectswere
perceivedpositionof thesegroupsin relation- again given the original attitude scale,
ship to the various attitude-content areas additional referencegroup data were ob-
studied. tained,and each subjectwas asked to state
Of the originalpool of subjects,204 were anonymouslyhis or her experiencewith
later selected to participatein a "voting marijuana.Responsesto thisquestionyielded
study."Carewas takenthatthelatternot be percentagesof subjects admittingvarious
associated with the originalquestionnaire levels of marijuanause similar to those
study. The 204 subjects selected for the obtained in an earlierstudy on the same
experimental phase of the researchincluded campus(L. DeFleurandGarrett, 1970).
72 fromeach of thefirstandfourth quartiles In overview, thevariables ofconcernto the
on theattitudescale and 30 fromeachofthe presentreportand theiroperationalformcan
two middlequartiles.The former representedbe summarized as follows:
those subjects having the most clear-cut Attitudeand BehaviorIntentions.A
attitudestowardmarijuanalegalizationand sixteen-itemLikert scale was developed
use, while the subjects from the middle measuring generalattitudetowardmarijuana.
quartileswereselectedprimarily forcompari- Respondents wereassigneda valueof 1 ifthey
sonpurposes.As notedby AcockandDeFleur scored in the firstor second (favorable)
(1972), thisrelativede-emphasisofthemiddle quartiles on the scale and a value of 0
quartileswas foundto have introducedno otherwise ?
biasesintotheanalysis. Thoughnot directly intendedat thetime,
The votingstudyconsistedofinviting each the scale includedtwo itemswhichcan be
of the 204 subjectsto the social research seen to approximateFishbein's behavior
laboratorywhere they were eventuallyre- intentions. These two itemswereas follows:
quired to participatein a behavioralact
(eithervotingor signing a petitionin support l 0The primary reasonfordichotomizing scores
of or in oppositionto marijuana legalization). on the attitudescale and treatingit as a dummy
The act could occur under variousexperi- variablein the analysiswas to maketheprocedure
mental conditionsincludingtwo levels of analogous to that used by Acock and DeFleur
(1972). The decomposition of data havingordinal
disclosure,twolevelsof congruence, andthree properties intodummyvariables in orderto employ
levelsof socialparticipation.9The two levels correlationand regressiontechniqueshas now
of disclosuresare the onlyones salientto the becomea widelyacceptedpractice(see,forexample,
presentanalysis.Disclosurewas variedby Boyle, 1970; Wulbert,1971; and Miller and
Erickson,1974). It shouldbe noted,however, that
makingthe act eitherpublic in nature or thismethodology usesonlynominalinformation in
strictlyprivate. In the formercondition, theanalysis.Usingthefullrangeof thescalewould
subjects were requested to sign either a likelyresultin highercoefficients becausemoreof
favorableor an unfavorablepetition on theinformation wouldbe used.
marijuanalegalizationwhich,theyweretold, scales Othersof thevariables weretreatedas five-point
in theanalysis(forexample,theBI measures).
would be made public throughthe news- Thiswasdonebecauseofa neutralalternative on the
media. In the private condition,subjects scales. Dichotomizationof the data on these
registereda vote for or againstmarijuana variableswould have entailedutilizingsuch pro-
ceduresas -1, 0, +1 categoriesor treating
neutral
responsesas missingdata.Bothof thesetechniques
were tried: the formerresultedin practicallyno
9The operationalization of the conceptsof changeinthecoefficients;
thelatter(treating
neutral
congruence and social participation
is discussedin as missing
responses data) resulted
in slightly
higher
Frideres
eta. (1971). correlations
betweenthevariables.

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
6 SOCIOMETRY
1) "I would signa petitionto do awaywith so faras theirattitudesand behaviortoward
on the use of marijuana," marijuanawere concerned.The measureof
legal restrictions
and 2) "I would not trymarijuanaevenif I normativebelief utilizedwas the perceived
werecertainI wouldnot getcaught."It will directionof influence(eitherpro- or anti-
be noted thatthesetwo itemsrelatedirectly marijuanause) of thepersonor groupranked
to the two measuresof overtbehaviorthat firstby the subject. For the analysis,the
were obtainedat a laterpointin the study. variablewas scoredas 0 iftheperceived norms
Therefore,they will be treatedas behavior of the most significant other were unfavorable
intentionstatementsand will be analyzed to marijuanause and as 1 if perceivednorms
separately. Each of theseitemswas scoredon werefavorable to use.
a five-point scaleranging fromstrongly agree The second measureof normative beliefs
to strongly disagree.1
1 was obtained in order to getan indicator that
NormativeBeliefs. The firstmeasureof relatedmoredirectly to thesecondbehavioral
normative beliefsapproximates thatreported measure (reported experience with mari-
by Acock and DeFleur (1972). They em- juana). In theanalysis,theinitialNB measure
ployed a measureof subject'sperceptionof (perceptionsof normsrelatingto marijuana
familyand peer attitudestowardmarijuana legalization)is used for the firstbehavioral
legalization.The data allow a somewhat more measure(votingor signing a petitionin favor
specificmeasurement of this variable,how- of or opposed to marijuana legalization),
ever. Each subject was asked to rankfive whilethe secondNB measureis usedwiththe
reference groupsin termsof theirimportance indicatormeasuredin thefollow-up interview.
in influencing hisorherbehavior:1) myclose OvertBehavior.The twomeasuresof overt
friendshere at school, 2) my close friends behavior employed in this analysisrelate
back home,3) myimmediate family,4) other directlyto the two behaviorintention items.
relativesand kin, and 5) facultymembersI In thelaboratory studysubjectswererequired
know. Subjectswerethenaskedto rateon a to eithervote or signa petitionopposedto or
five-pointscale how each of these groups in favorof marijuanalegalization.As noted,
would feel about marijuanalegalization.The thebehavioral measures werevariedsomewhat
measureof normative beliefsemployedhere for the public and privateconditions.In the
was obtainedfromascertaining each subject's former, petitionsweresigned;in thelatter,a
perceptionof the normative positionof the privateballotwas checked.An equal number
reference grouphe or shepersonally rankedas of respondents participated in eachof thetwo
most important.(Subject behaviorwas also conditions. Analysis revealedthattheBI item
compared with generalnormative perceptions; employed in this instance predicted the two
i.e., "how do most people in the American similar types of behavior equally well.
societygenerally feelaboutmarijuana legaliza- Therefore, theyhavebeen treatedtogether in
tion?"Thisyieldedsignificantly lowerpredic- thediscussionwhichfollows.In thefollow-up
tiveequations).Scoreson thisvariablecould interview, subjectsindicatedlevelof personal
rangefrom1 to 5 or fromstrongly favorable experiencewith marijuana. For both be-
to stronglyunfavorable. haviors,a score of 0 was assignedto an
The measureof normative beliefobtained unfavorable behaviorresponse(eitherregister-
in the follow-up interviewwas closelysimilar ing a negative vote or petitionin the lab or
to the above. Subjectswere asked if there reporting not to have used marijuanain the
were any persons or groups who had follow-upinterview)and a score of 1 was
influencedthemin eithera favorableor an assignedto a favorableresponse(a positive
unfavorablemannertowardmarijuanause. vote or petitionor admissionof personal
Theywerethenaskedwhichof thegroupsor use).'2
personsmentioned had themostinfluence in
1'In' an effortto utilizethe completeFishbein
l As noted, the behavior intention statements model,a measureof motivation to complywas also
were initially developed as items in the 16-item tested.Theoriginal completed
opinionquestionnaire
Likert scale. For this analysis, these items were by all subjectsincludedscalesto measuresubject's
extractedfromthe initial scale for the computation need for social approvaland inner-other directed-
of the A-BI relationshipsin orderthattheirinclusion ness. In this analysisit was assumedthatpersons
inflatethe relationshipsobserved.
not artificially characterizedby high need for approval or by other

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
TABLE 1 7

for Lab Behavior (PetitionSigningand Voting)


CorrelationCoefficients
and Related BehaviorIntention(N = 204ff

CorrelationCoefficients CorrelationCoefficients Multiple Multiple


Correlation Correlation Correlation
ATT-BEH NB-BEH ATT-BI NB-BI on BEH on BI BI-B

.539 .369 .638 .522 .545 .667 .448


Note: ATT = attitude toward marijuana; BEH = behavior; NB = normativebeliefs; BI = behaviorinten-
tions
aall correlationcoefficientsare significant
at the .001 level.

Summary ofFindings betweenthe two independent variables.This


Table 1 presents relevant correlation is in fact the case (r = .873). Again,the
forthe initial(laboratory)phase relationships
coefficients betweenthe independentvari-
of the study. It can be seen that in the ables and the BI itemare somewhatstronger
zero-ordercorrelationcoefficients, attitude than for the behavioral measureand in this
accountsfor approximately threetimesthe instance the BI-B relationshipis quite a bit
variationin behaviorcommonlyobservedin stronger than for the A-B relationship.
bivariateattitude-behavior studies(as summar- Discussion
ized by Wicker,1969). The normative belief
variable,thoughsignificantly relatedto the Several general commentscan be made
laboratorybehavior,does not add greatlyto about the relationships observedin the two
the amount of explained variationwhen tables:
combinedwith attitude(R = .545; R2 = 1) The strength of therelationship between
.294). attitudeand behavioris strongerthan that
Both attitudeand normativebeliefsare betweenthe specificBI itemand behaviorin
more stronglycorrelatedwith the behavior thelaboratory studybut not in thefollow-up
intentionitem("I wouldsigna petitionto do study. That is, in the firstinstancethe
away with legal restrictions on the use of attitudescale gives a betterindicationof
marijuana")thanwithactualbehavior.Inter- probablebehaviorthan does a specificbut
estingly,the specificBI itemis not strongly singleitem.The assumptionthatthisshould
relatedto actualbehavior.In fact,thegeneral generallyhold truehas been a widelyused
attitudescale offersbetterpredictionof the argumentfor the employmentof attitude
overtact thandoestheBI item. scales as opposed to one or two item
Table 2 presents the same statistical statements of beliefor behaviorintention.
summaryfor the secondbehaviormeasure- There may be severalreasonswhy this
self-report of personalexperiencewithmari- observation does not hold for the second
juana. In this instance,the generalattitude- behavioral measure. In the firstplace, the
overtbehaviorrelationship is slightlylower behavior (use or nonuse of marijuana) had,in
than in the firstinstance. However,the all probability, occurred prior to the attitude
correlationbetween normativebeliefs and measurement(only the self-report occurs
behaviorincreases.The slightimprovement in after), resulting in a possible commitment
explainedvariationfromthe multiple(R = effect.Or, more likely,the BI item in the
.538, R2 = .289) over the zero-order originalscale simplybecameanothermeasure
relationships suggestsstrongintercorrelationofself-reported behaviorforthesubjects.
A secondpossibility, followingTittleand
Hill (1967), is that the second behavioral
directedness would be more highlymotivatedto
complywithperceived normative That
expectations.
measurerepresents an actionalternative that
is, motivationto complywas viewedas a general is more likely to be found within the
responsetendency oftheindividual. individual's
or characteristic commonrangeof experiences. As
Consistent with Fishbein'srecurrent the a result,one would expect a stronger
findings, BI-B
MC variabledid notincreasethelevelof prediction relationship thanforbehaviorsthatrepresent
attainedby normativebeliefsalone. Given the
consistency it would perhapsbe
of this finding, what Tittle and Hill referto as moreunusual
advisableto dropthiscomponent fromthe model behaviorcontextsor options.Theyfeelthat,
altogether. generally, laboratorymeasuresfitunderthe

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
8 TABLE 2

CorrelationCoefficients
forFollow-upStudy (Self-Reportof Marijuana Use)
and Related BehaviorIntentions(N= 59)a

CorrelationCoefficients CorrelationCoefficients Multiple Multiple


Correlation Correlation Correlation
ATT-BEH NB-BEH ATT-BI NB-BI on BEH on BI BI-B

.511 .527 .645 .618 .538 .655 .753


Note: ATT = attitude toward marijuana; BEH = behavior; NB = normativebeliefs; BI = behaviorinten-
tions
aall correlationcoefficientsare significant
at the .001 level.

latter categorizationwhile non-laboratorybecomes severely restricted.By defining


measuresof behaviorare morelikelyto fall attitudesmore generally, and when dealing
within the individual'scommon range of with actual behaviorsas opposed to state-
experiences. mentsof intentionto behave,we havecome
2) Related to the latterpoint,normative up with coefficients that can be considered
beliefsbecome strongerpredictorsof both quite high. It should be noted that the
behaviorand behaviorintentionsfor the measuresof overtbehavioremployedin this
secondbehavioralmeasure.Again,thisseems researchwere both obtainedseveralweeks
reasonable.Normsrelatingto marijuanause aftertheinitialmeasurement of attitudesand
are probablymuch more salientthanthose behaviorintentions. Thislapsein timemaybe
relatingto votingfor or againstmarijuana a criticalfactorin causingthe moregeneral
legalization.Therefore,subjects would be attitudemeasureto be the betterpredictor
more likely to take perceivednormsinto variable.In most of the researchemploying
account when decidingto use or not use the basic Fishbeinmodel,behavioris mea-
marijuanathan when votingfor or against suredverysoon afterthemeasureofbehavior
marijuana legalization. intentions.
3) Usinga general(and traditional)
attitude At the sametime,our data suggestcertain
measureand measuresof actualbehavior.we empiricalquestions concerningthe model
foundrelatively strongrelationships.Though whichsupplement thoseconceptualquestions
the coefficientsare not as large as those discussedearlier.The inclusionof normative
frequently reportedby Fishbein,theyseemto beliefs in this particularresearchdid not
have much more generalapplicabilityand greatly improvetheprediction of behavioror
suggestthatthe modelneednot be confined behaviorintentionin eitherthelab or in the
to narrowly definedattitudestowardspecific field over that obtainedwith the attitude
acts nor to behaviorintentionsratherthan measure alone. By the same token, the
behavior.It shouldbe notedfurther thatthis oppositecouldalso be argued:theadditionof
report does not take into consideration attitude to the model adds little to the
interaction effectsas do Acock and DeFleur predictionobtainedby the NB variable.The
(1972). Their findings indicatethatpredict- factthatprediction ofbehavioral outcomesin
abilityfromthemodelwouldbe evenhigher the present instance is not particularly
whenincluding suchinteractions. improvedby inclusion of both variables
In conclusion,our analysissuggeststhat appears to be a functionof the fact that
the general type of model employedby personalattitudeand normative beliefswere
Fishbeinand his colleaguesmay have more highlyinterrelated. In fact,as notedearlier,
generalapplicability thanis evidentfromthe thezero-order correlation of thesevariablesis
studiesin whichit has been utilizedto this .87, suggesting that the two majorindepen-
point.Our finding, thata generalmeasureof dentvariablesin thisstudymeasurebasically
attitudesdoes equally as well in predicting the same thing.1 3 Whilethe model implies
behavioras the more specificmeasuresof
behaviorintention, suggestsgreaterrobustness
of the model. As suggestedearlier, if
13 argument is furthersupported whenwe
attemptto "sort-out"theindependent ofthe
effects
prediction of behavioris limitedto situations twovariables.On theone hand,themultiple R adds
whereveryspecificBI itemsmustbe written verylittleto what is obtainedusingone of the
foreach possibleact,theabilityto generalize independent alone.On theother,first
variables order

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
ATTITUDES,BEHAVIORAND BEHAVIORINTENTIONS 9
relativeindependence, we would arguethat Albrecht,Stan L., MelvinL. DeFleurand Lyle G.
such a presumption may not be necessary. Warner
Attitudesare learnedlargelythroughinter- 1972 amination
"Attitude-behavior relationships:A reex-
of the postulateof contingent
actionand associationwithsignificantothers, consistency." PacificSociologicalReview
who mayalso providetheimportant referent 15:149-168.
pointsin assessingnormativebeliefs.It will Boyle,RichardP.
probably be true that the independent 1970 "Path analysisand ordinaldata." Ameri-
contributionofNB willvaryfromsituationto canJournal of Sociology75:461-480.
A. R.
Carlson,
situationdepending,in part,on the congru- 1968 "The relationships betweena behavioral
ence betweenindividualattitudesandpercep- intention,attitudetowardthe behavior
tionofsocialnorms. andnormative beliefsaboutthebehavior."
Finally,the suggestionby DeFleur that Unpublished doctoraldissertation,Univer-
sityofIllinois.
interactionsbetween componentsof the Darroch,R. K.
model be consideredseems critical. The 1971 "Attitudinal variables andperceivedgroup
influenceoftheNB variableis ratherdifferent normsas predictors of behavioralinten-
forthe two behavioralmeasuresemployedin tions and behaviorin the signingof
photographic releases."Unpublished doc-
thisresearch.Thissuggeststhatperceptions
of toraldissertation,UniversityofIllinois.
social normsmay interactdifferently with DeFleur,LoisandGeraldGarrett
individualattitudesdependingon thesalience 1970 "Dimensionsof marijuanause in a
of one or the other to the particular land-grant university." Journalof Counsel-
behavioralsituation.Researchthatwill take DeFleur,M.ingPsychology 17:468-476.
L. andF. R. Westie
such considerationsintoaccountwouldseem 1958 "Verbal attitudesand overt acts: An
to offergreaterpotentialforimproving our experiment on the salienceof attitudes."
understanding of the ubiquitousattitude-be- American SociologicalReview 33:667-
haviorproblem. 673.
"Attitudeas a scientific
1963 concept."Social
Forces42:17-31.
coefficients show significantly lower relationships DeFriese,G. andW.S. Ford
wheneitherthe effectof attitudeor the effectof 1968 "Open-occupancy: Whatwhitessay,what
normative beliefsis controlledfor.For example, theydo." Transaction 5:53-56.
r,3 .2 = .13 andr231 =.19 when X Attitude, X Deutscher,I.
Normative Beliefs,and X3 = Behavior(self-report 1966 "Wordsand deeds: Social science and
marijuana use). social policy.'Social Problems13:235-
254.
REFERENCES 1969 "Lookingbackward:Case studieson the
progressof methodologyin social re-
Acock,AlanC. andMelvinL. DeFleur search."American Sociologist4:35-41.
1972 "A configuration approachto contingent 1973 What We Say/WhatWe Do. Glenview,
consistency in the attitude-behaviorrela- Illinois:Scott,Foresman andCompany.
tionship."AmericanSociologicalReview Ehrlich, H.
34:714-726. 1969 "Attitudes, behaviorand the intervening
Ajzen,I. variables."American Sociologist4:29-34.
1971 "Attitudinal vs. normativemessages:An Fendrich, J.
investigation of the differentialeffects
of 1967 "A studyof theassociationamongverbal
persuasive communications on behavior." attitudes, commitment, andovertbehavior
Sociometry 34:263-280. in differentexperimentalsituations."
Ajzen,I. andM. Fishbein SocialForces45:347-355.
1969 "The predictionof behavioral M.
intentions Fishbein,
in a choicesituation."Journalof Experi- 1963 "An investigation of the relationships
mentalSocialPsychology 6:400416. betweenbeliefsabout an objectand the
1970 "The prediction ofbehaviorfromattitudi- attitude toward that object." Human
nal and normative variables."Journalof Relations16:233-240.
Personality and Social Psychology 1967 Attitudeand the Prediction of Behavior.
6:466-487. In M. Fishbein(ed.) Readingsin Attitude
1972 "Attitudes and normative beliefsas factors TheoryandMeasurement. NewYork:Wiley.
influencing behavioral intentions." Journal Wiley.
ofPersonality andSocialPsychology 21:1-9. 1973 "The prediction of behaviorsfromattitu-
dinal variables."In D. C. Mortensen and
1973 "Attitudinaland normative variablesas K. K. Sereno(eds.),Advances in Communi-
predictors of specificbehaviors." Journal cation Research.New York: Harperand
of Personalityand Social Psychology Row.
27:41-57. J.S., L. G. Warner
Frideres, andS. L. Albrecht

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions
10 SOCIOMETRY
1971 "The impactof social constraints on the attitude-behaviordiscrepancies."Journal
relationshipbetween attitudesand be- of Personalityand Social Psychology
havior."SocialForces50:102-112. 24:225-236.
Hornik, J.A. Tittle,C. andR. Hill
1970 "Two approachesto individual differences 1967 "Attitudemeasurement of
and prediction
in cooperativebehaviorin an expanded behavior:An evaluation ofconditionsand
prisoner'sdilemmagame." Unpublished measurementtechniques."Sociometry
master's levelpaper,UniversityofIllinois. 36:199-213.
LaPiere,R. Warner, L. G. andM. L. DeFleur
1934 "Attitudes vs. action." Social Forces 1969 "Attitudeas an interactional concept:
13:230-237. Social constraintand social distanceas
Linn,L. interveningvariablesbetween attitudes
1965 "Verbalattitudesand overtbehavior:A andactions."American Review
Sociological
study of racial discrimination." Social 34:153-169.
Forces43:353-364. Wicker, A. W.
Miller,
Jerry L. L. andMaynard Erickson 1969 "Attitudes vs.actions:Therelationshipof
1974 "On dummyvariableregression analysis." verbaland overtbehaviorresponsesto
Sociological Methods and Research attitudeobjects."Journalof Social Issues
2:409-430. 25:41-78.
Minard,R. D. Wulbert, Roland
1952 "Race relationsin the Pocahontascoal 1971 "Multivariateanalysis of dichotomous
fields."Journal ofSocialIssues8:29-44. variables:A generalmethod."Multivariate
Schwartz,ShalomH. andR. C. Tessler Behavioral Research6:215-232.
1972 "A testof a modelforreducing measured

MANUSCRIPTS FOR THE


ASA ROSE SOCIOLOGY SERIES
Manuscripts( 100 to 300 typedpages) are solicitedforpublicationin
the ASA Arnold and Caroline Rose MonographSeries. The Series
welcomes a varietyof types of sociological work-qualitative or
quantitativeempirical studies, and theoretical or methodological
treatises.An authorshouldsubmitthree copies of a manuscript for
considerationto the Series Editor,ProfessorIda HarperSimpson,De-
partment of Sociology,DukeUniversity,Durham, NorthCarolina27706.

This content downloaded from 62.122.78.91 on Wed, 25 Jun 2014 10:39:28 AM


All use subject to JSTOR Terms and Conditions

You might also like