Download - Moderation of cognition-intention and cognition-behaviour relations: A meta-analysis of properties of variables from the theory of planned behaviour

Moderation of cognitionndashintention andcognitionndashbehaviour relations A meta-analysisof properties of variables from the theory ofplanned behaviour

Richard Cooke and Paschal SheeranUniversity of Sheffield UK

Meta-analysis was used to quantify the moderating effects of seven propertiesof cognitionsmdashaccessibility temporal stability direct experience involvementcertainty ambivalence and affective-cognitive consistencymdashon cognitionndashintentionand cognitionndashbehaviour relations Literature searches revealed 44 studies that couldbe included in the review Findings showed that all of the properties except involve-ment moderated attitudendashbehaviour consistency Similarly all relevant moderatorsimproved the consistency between intentions and behaviour Temporal stabilitymoderated PBCndashbehaviour relations certainty moderated subjective normndashintentionrelations and ambivalence certainty and involvement all moderated attitudendashintention relations Overall temporal stability appeared to be the strongestmoderator of cognitionndashbehaviour relations

The theory of planned behaviour (TPB Ajzen 1985 1991) is probably the dominantaccount of the relationship between cognitions and behaviour in social psychologyThe theory posits that the proximal determinant of an individualrsquos behaviour is hisherintention to perform it Intentions (eg lsquoI intend to exercise at least three times aweekrsquo) reflect how hard people are willing to try to achieve their goals and alsosummarize individualsrsquo motivation to perform a behaviour (Ajzen 1991 p 181) Inten-tions are predicted by attitudes subjective norms and perceived behavioural control(PBC) Attitudes are an individualrsquos positive or negative evaluation of performing thebehaviour (eg lsquoFor me to exercise at least three times a week would be goodbadrsquo)Subjective norms are an individualrsquos beliefs about what significant others think sheshould do (eg lsquoMost people who are important to me think that I should exercise atleast three times a weekrsquo) PBC reflects perceptions of the ease or difficulty of behav-ioural performance (eg lsquoFor me to exercise at least three times a week would beeasydifficultrsquo) PBC can also directly predict behaviour when PBC accurately reflects

Correspondence should be addressed to Richard Cooke Centre for Research in Human Behaviour School ofSocial Science and Law Collegiate Crescent Campus Sheffield Hallam University Sheffield S10 2BP UK (e-mailrichardcookeshuacuk)

British Journal of Social Psychology (2004) 43 159ndash186copy 2004 The British Psychological Society

wwwbpsorguk

159

the personrsquos actual control over behavioural performance (Sheeran Trafimow ampArmitage 2003)

Meta-analyses indicate that the TPB provides a good explanation of a wide range ofbehaviours For example attitudes subjective norms and PBC account for 39ndash42 ofthe variance in intentions (Armitage amp Conner 2001 Godin amp Kok 1996 Sheeran ampTaylor 1999) while intentions and PBC predict between 28 and 34 of the variance inbehaviour (Armitage amp Conner 2001 Godin amp Kok 1996 Trafimow Sheeran Conneramp Finlay 2002) Notwithstanding the success of the TPB there remains a substantialproportion of variance in intentionsbehaviour that is not explained by TPB variables(Sheeran 2002)

One line of research designed to increase the predictive validity of the TPB has beento examine variables that might moderate cognitionndashintention and cognitionndashbehaviour relations A moderator variable affects the relationship between two othervariables (cf Baron amp Kenny 1986) For example the relationship between exerciseattitudes and exercise behaviour may be moderated by the certainty of participantsrsquoattitudes such that participants who possess highly certain attitudes exhibitstronger attitudendashbehaviour relations than participants who are less certain of theirattitudes The present study focuses on seven properties classified by Krosnick andPetty (1995) as relating to defining features of attitude strength (temporal stability)aspects of attitude structure (accessibility affective-cognitive consistency ambiva-lence) subjective beliefs about attitudes (certainty involvement) or processes ofattitude formation (direct experience)

Properties of cognitions as moderator variables

Krosnick and Petty (1995) suggested that all seven properties are likely to improveattitudendashbehaviour consistency though they also note that these properties are oftencorrelated with one another This could indicate that these properties improveattitudendashbehaviour consistency because they reflect a single underlying constructHowever contrary to this hypothesis research by Krosnick Boninger Chuang Berentand Carnot (1993) suggested that properties of attitudes are conceptually distinctUsing confirmatory factor analysis Krosnick et al (1993) found that a single-factormodel provided an inadequate description of the structure of ten attitudinal propertiesdespite significant correlations between the properties Exploratory factor analysesfound the best fit for a five-factor model with no more than three properties loadingon any one factor (see also Bassili 1996 Erber Hodges amp Wilson 1995 Prislin 1996)Thus although properties of attitudes may be correlated with one another they areconceptually distinct constructs

Researchers have also employed these properties as moderators of relationswithin the TPB Studies have demonstrated that attitude strength indices canmoderate attitudendashintention relations (Armitage amp Conner 2000 Budd amp Spencer1984) and can be adapted to moderate subjective normndashintention (Nederhof 1989Trafimow 1994) intentionndashbehaviour (Bassili 1995 Sheeran Orbell amp Trafimow1999) and PBCndashbehaviour (Conner Sheeran Norman amp Armitage 2000 Doll ampAjzen 1992) relations The aim of the present study is to conduct the first quantitativereview of the impact of the seven moderator variables outlined above on cognitionndashintention andor cognitionndashbehaviour relations In the next section we define eachof the properties and describe previous research that tested moderation by theseproperties

160 Richard Cooke and Paschal Sheeran

AccessibilityAccessibility refers to the strength of the association in memory between a cognition(eg attitude) and the object of that cognition (eg exercise behaviour) and is usuallymeasured by the latency between presentation of a question designed to measurethe cognition and the participantrsquos response (Fazio 1995) Alternatively some studieshave manipulated accessibility by asking one group of participants to repeatedlyexpress their attitudes and another group of participants to express their attitudesonce this procedure is assumed to make the attitudes of the repeated expressiongroup more accessible compared with those of the single expression group Studieshave shown that both indices of accessibility can moderate cognitionndashbehaviourrelations For example Fazio Chen McDonel and Sherman (1982) found thatparticipants who repeatedly expressed their attitudes towards puzzle types weremore likely to act in accordance with these attitudes than were participants whoexpressed their attitudes once Using a response latency measure Fazio and Williams(1986) found that participants with highly accessible attitudes possessed greaterattitudendashvoting behaviour consistency than participants with less accessible attitudesin the 1984 US presidential election Other research has supported and extendedthese findings (Bassili 1995 Fazio Powell amp Williams 1989 Kokkinaki amp Lunt1997)

However accessibility does not always moderate cognitionndashbehaviour relations Forexample Doll and Ajzen (1992) examined the moderating effects of accessibility andtemporal stability on relations between TPB variables and video game play They foundthat accessibility did not moderate attitudendashbehaviour intentionndashbehaviour or PBCndashbehaviour relations whereas temporal stability moderated all three relationships(see also Cooke amp Sheeran 2004a) Consequently Doll and Ajzen argued thattemporal stability is a more effective moderator of cognitionndashbehaviour relations thanis accessibility

Temporal stabilityTemporal stability can be defined as the extent to which cognitions remain consistentover time (Sheeran et al 1999) and is generally measured by the within-participantscorrelation between cognition items taken at two different timepoints According toKrosnick and Petty (1995) temporal stability is a defining feature of strong attitudesThe other defining feature of attitude strength is impact and there is evidence thattemporal stability may be antecedent to behavioural impact (Sheeran et al 1999)Temporal stability should moderate cognitionndashbehaviour relations because as Ajzen(1996) argued lsquoto obtain accurate prediction of behaviour intentions and perceptionsof behavioural control must remain reasonably stable over time until the behaviour isperformedrsquo (p 389) If a participantrsquos intention or PBC changes before their behaviouris measured then intention or PBC may not accurately predict behaviour Support forAjzenrsquos hypothesis came from Conner et al (2000) they showed that for both afrequently performed behaviour (eating a low-fat diet) and an infrequently performedbehaviour (attending a health check-up) temporal stability moderated intentionndashbehaviour relations such that more stable intentions were better predictors of behav-iour than were to less stable intentions Similar findings were obtained in other studiesthat tested the moderating effects of temporal stability on attitudendashbehaviour andPBCndashbehaviour relations (Cooke amp Sheeran 2001a Davidson amp Jaccard 1979 Doll ampAjzen 1992)

Moderators of the TPB 161

Direct experienceDirect experience refers to whether or not participants have performed a behaviourprior to measurement of cognitions and behaviour in a particular experiment Indirectexperience on the other hand is characterized by activities such as reading aboutbehavioural performance or observing someone else performing the behaviour priorto measurement of cognitions and behaviour Hence direct experience is usuallymanipulated by giving one group of participants practice with the focal behaviourwhereas the control group read about or observe the behaviour Research suggeststhat the impact of direct experience on cognitionndashbehaviour relations may bemediated by other properties of cognitions For example Fazio (1995) argued thataccessibility mediates the impact of direct experience on cognitionndashbehaviour con-sistency and research has shown that direct experience produces more accessiblecognitions which in turn provide greater prediction of behaviour (eg Fazio et al1982) In contrast Ajzen and Fishbein (2001) proposed that temporal stability medi-ates the influence of direct experience on behaviour They argued that cognitionsformed through direct experience provide participants with relevant informationabout the consequences of performing the behaviour Therefore participants are likelyto use these cognitions again during subsequent decisions to perform the behaviourThe upshot is that cognitions formed by participants with direct experience are likelyto be more stable than those held by participants who have not performed the behav-iour (whose cognitions are likely to change when the behaviour is actually experi-enced) Whichever interpretation of direct experience is accurate findings show thatdirect experience tends to strengthen cognitionndashbehaviour consistency (see Fazio ampZanna 1978b Millar amp Millar 1996 Regan amp Fazio 1977)

InvolvementAn issue is involving when it is perceived as important and personally relevant (egKokkinaki amp Lunt 1997) Thus the prediction is that participants who are highlyinvolved with an issue should possess greater attitudendashintention consistency than lessinvolved participants Evidence to support this view was obtained by Petty Cacioppoand Schumann (1983) who found that participants who were more involved with aproduct showed greater attitudendashintention consistency and were less persuaded byweak arguments for the product compared with participants who were less involvedSimilarly Verplanken (1989) showed that participants who were highly involved in theissue of nuclear energy possessed stronger attitudendashintention relations than partici-pants who were less involved (see also Nederhof 1989) Petty Haugtvedt and Smith(1995) suggested that involvement moderates attitudendashintention consistency becauseit produces attitudes that are based on greater elaboration of relevant informationwhich in turn produces attitudes that are more accessible certain and based on moreknowledge Studies by Kokkinaki and Lunt (1997 1999) and Lavine Borgida andSullivan (2000) showed that high levels of involvement with an object produces moreaccessible attitudes

CertaintyCertainty is usually measured using a single item that asks participants how certainthey feel about their cognitions (eg lsquoMy intention to vote for X is certainuncertainrsquo)Fazio and Zanna (1978a) found that participants who were more certain about theirattitudes possessed stronger relations between attitudes and volunteering behaviour


than participants who were less certain Trafimow (1994) found that certaintymoderated subjective normndashintention relations participants who were highly certainabout their subjective norm in relation to using a condom possessed greater subjectivenormndashintention consistency than participants who were less certain of their subjectivenorm Similarly Bassili (1993) found that certainty moderated the relationship betweenintention and voting behaviour Bassili (1996) suggested that certainty has a specificityof meaning (ie to be certain is to be unbudgeable on an issue) which may explain whycertainty is a successful moderator of cognitionndashintention and cognitionndashbehaviourassociations

AmbivalenceAttitudes are usually conceived as unidimensional bipolar constructs and participantsare assumed to possess either a positive neutral or negative attitude towards anattitude object However Kaplan (1972) pointed out that individuals could hold bothpositive and negative attitudes towards the same attitude object For example a personmay hold both positive and negative attitudes towards eating a low-fat diet (eglsquoEating a low-fat diet is healthyrsquo vs lsquoEating a low fat diet is unpleasantrsquo) To calculateambivalence scores most researchers use the formula provided by Thompson Zannaand Griffin (1995) that captures both the similarity and intensity of participantsrsquopositive and negative attitudes

Sparks Conner James Shepherd and Povey (2001) argued that high levels ofambivalence reduce attitudendashintention and attitudendashbehaviour consistency becauseparticipantsrsquo attitudes are in conflict this conflict creates less stable attitudes thatoffer poor prediction of intentionsbehaviour For example Armitage and Conner(2000) found that participants with less ambivalent attitudes towards eating a low-fatdiet possessed greater attitudendashintention and attitudendashbehaviour consistency thanparticipants high in ambivalence However not all research has confirmed the moder-ating impact of ambivalence outlined above Jonas Diehl and Bromer (1997) foundthat higher levels of ambivalence were associated with greater attitudendashintentionconsistency Jonas et al claim that higher ambivalence leads to more systematicprocessing of available information which produces attitudes that are more predictiveof intentions (cf Petty and Caciopporsquos [1986] elaboration likelihood model ofpersuasion)

Affective-cognitive consistencyRosenberg (1960 1968) proposed that attitudes possess both an affective and a cogni-tive component the affective component refers to a participantrsquos feelings about per-forming a behaviour whereas the cognitive component reflects a participantrsquosthoughts about the behavioural performance Researchers compute the discrepancybetween affective and cognitive attitude measures to calculate affective-cognitiveconsistency Rosenberg (1968) stated that affective-cognitive consistency shouldstrengthen the attitudendashbehaviour relationship because lsquowhen the affective andcognitive components of an attitude are mutually consistent the attitude is in a stablestatersquo (p 75) Thus individuals with high affective-cognitive consistency wereexpected to exhibit more stable attitudes that may be better predictors of behaviourthan the attitudes of individuals with low affective-cognitive consistency Norman


(1975) investigated this hypothesis in a study of studentsrsquo volunteering behaviour andfound that participants with high affective-cognitive consistency possessed strongerattitudendashbehaviour relations than participants with low affective-cognitive consistencyHowever Fazio and Zanna (1978a) in a replication of Normanrsquos study found nomoderating effect of affective-cognitive consistency on attitudendashvolunteer behaviourrelations

The present study

The present study examines the impact of seven properties of cognitions (accessibilitytemporal stability direct experience involvement certainty ambivalence andaffective-cognitive consistency) as moderators of five relationships in the TPB namelyattitudendashbehaviour intentionndashbehaviour PBCndashbehaviour attitudendashintention and sub-jective normndashintention relations The rationale for the current study is that previoustests of moderation by these properties have produced contradictory findings Thus itis unclear how well these properties moderate cognitionndashintention and cognitionndashbehaviour relations across studies The main aim of the present study is to provide thefirst quantitative review of the properties of cognitions as moderators of cognitionndashintention and cognitionndashbehaviour relations A second aim of this study is to test thecomparative strength of the properties in order to identify which moderator(s) bestenhance cognitionndashintention and cognitionndashbehaviour associations

Method

Sample of StudiesSeveral methods were used to generate the sample of studies (a) computerizedsearches of social scientific databases (Dissertations Abstracts Online Index to ThesesPsychLIT Social Science Citation Index and Web of Science) from January 1981 to thetime of writing (November 2002) (b) reference lists in each article were evaluated forinclusion and (c) the authors of published articles were contacted and requests weremade for unpublished studies and studies in press

For inclusion in the review a bivariate statistical relationship between cognitionsand intention (or behaviour) for participants classified as high or low on the moderatorvariable had to be retrievable from studies Where studies did not include relevantstatistics the authors of the study were contacted and requests were made for bivariateassociations Using this inclusion criterion a total of 44 tests of cognitionndashintentionbehaviour associations were found An asterisk in the reference list precedes the 44studies that yielded the 94 effect sizes These 44 studies included 9 unpublished papers(Conner Povey amp Sparks 1995 Cooke amp Sheeran 2001 2004a 2004b Godin ampConner 2002 Godin Conner Gagnon amp Lambert in press Huiuk 1995 Rennier1989 Sheeran 1999) Table 1 presents the characteristics and effect sizes obtainedfrom the 45 studies included in the review Thirty-three studies (73) reportedcognitionndashbehaviour correlations eight studies reported cognitionndashintention relations(18) and five studies (11) reported both cognitionndashbehaviour and cognitionndashintention relations

Meta-analytic strategyThe effect size estimate employed here was a weighted average of the samplecorrelations r+ r+ describes the direction and strength of the relationship between






two variables with a range of minus10 to +10 Computing the weighted average effect sizerequires a transformation of the correlation from each relevant hypothesis into Fisherrsquos(1921) Z The following formula is then employed

where rzi=the Fisherrsquos Z transformation of the correlation from each study i andNi=number of participants in study i

In this way correlations based on larger samples receive greater weight than thosefrom smaller samples The average Z value is then back-transformed to give r+ (seeHedges amp Olkin 1985 Hunter Schmidt amp Jackson 1982)

Homogeneity analyses were conducted using the x2 statistic (Hunter et al 1982) todetermine whether variation among the correlations was greater than chance Thedegrees of freedom for the x2 test are k minus 1 where k is the number of independentcorrelations If x2 is non-significant then the correlations are homogeneous and theaverage weighted effect size r+ can be said to represent the population effect sizeTransformation of other statistics (eg t F values) to statistic r computation of theweighted average correlations and homogeneity analyses were all conducted usingSchwarzerrsquos (1988) Meta computer program

Multiple samples and multiple measuresWhere studies reported separate statistical tests for more than one sample then thecorrelation from each sample was used as the unit of analysis Where studies includedmore than one measure of cognitionndashintention or cognitionndashbehaviour relations (egattitudendashintention and subjective normndashintention relations see Budd amp Spencer 1984Nederhof 1989 Trafimow 1994) each correlation was treated as an independenttest of that relation Where studies contained multiple non-independent samples(Nederhof 1989 Norman 1975 Sparks Hedderley amp Shepherd 1992 Verplanken1989) we employed the conservative strategy of using the weighted average of thesample correlations and the smallest N in the analysis to determine the overall effectsize for that study Where studies had measured multiple moderator variables (Cooke ampSheeran 2004a 2004b Doll amp Ajzen 1992 Fazio amp Zanna 1978a 1978b Nederhof1989 Sheeran amp Abraham 2003) they were treated as independent tests of thesemoderators

Results

Two strategies were adopted to examine the moderating effects of the seven variablesFirst each moderator was treated as a dichotomous variable with participants definedas either lsquohighrsquo or lsquolowrsquo on the moderator This follows the convention in previousstudies The sample weighted average correlation r+ was calculated separately for thetwo groups and we used Fisherrsquos (1921) Z test for the comparison of independentcorrelations to assess the significance of the difference between effect sizes (one-tailedtests were employed because hypotheses were directional in all cases) This provided ameasure of the overall effectiveness of each variable as a moderator of cognitionndashintentionbehaviour relations Second each moderator was also analysed separatelyfor attitudendashbehaviour intentionndashbehaviour PBCndashbehaviour attitudendashintention andsubjective normndashintention relations where appropriate


AccessibilityWe hypothesized that participants with more accessible cognitions would demonstratestronger cognitionndashbehaviour associations Table 2 shows that overall accessibilitymoderated cognitionndashbehaviour relations Participants with more accessible cognitionspossessed stronger cognitionndashbehaviour relations (r+=60) than participants with lessaccessible cognitions (r+=52) Looking at specific cognitionndashbehaviour relationsaccessibility was a successful moderator of attitudendashbehaviour relations participantswho possessed highly accessible attitudes showed stronger attitudendashbehaviour consist-ency (r+=61) compared with participants with less accessible attitudes (r+=50) Inaddition participants with more accessible intentions demonstrated strongerintentionndashbehaviour consistency (r+=75) compared with participants with lessaccessible intentions (r+=62) In contrast accessibility did not significantly moderatePBCndashbehaviour relations (one-tailed p=17)

One factor that differentiated between studies employing accessibility as amoderator of attitudendashbehaviour consistency was the operationalization ofaccessibility Some studies manipulated accessibility whereas other studies measuredaccessibility Further analyses were conducted to discover whether this factorinfluenced how well accessibility moderates attitudendashbehaviour relations

Operationalization of accessibilityThree of the ten papers included in the review manipulated accessibility by asking onegroup to repeatedly express their attitudes and a separate group to express theirattitudes just once The remaining seven papers measured accessibility using responselatencies Therefore to compare the effects reported for the two different types ofaccessibility the weighted average of the sample correlations (for groups high and lowon accessibility) were calculated separately for repeated expression and responselatency studies Meta-analysis of the repeated expression studies showed there was amarginally (p lt 06) significant difference between the correlations for participantswith highly accessible attitudes (r+=56) and participants with less accessible attitudes(r+=41) There was also a significant difference between the correlations of partici-pants who exhibited faster (r+=64) compared with slower (r+=52) response latenciesThus the type of measure has a negligible effect on how well accessibility moderatesthe attitudendashbehaviour relation

Temporal stabilityTable 2 summarizes the findings for temporal stability Overall temporal stabilitymoderated cognitionndashbehaviour relations participants with more stable cognitionspossessed greater cognitionndashbehaviour consistency (r+=62) than participants with lessstable cognitions (r+=27) Temporal stability was also an effective moderator ofspecific relationships Participants with more stable attitudes possessed greaterattitudendashbehaviour consistency (r+=65) than participants with less stable attitudes(r+=minus01) participants with highly stable intentions showed stronger intentionndashbehaviour consistency (r+=67) than participants with less stable intentions (r+=30)and participants with more stable PBC demonstrated greater PBCndashbehaviourconsistency (r+=49) than participants with less stable PBC (r+=32)

Direct experienceDirect experience was predicted to increase the strength of cognitionndashbehaviourrelations Inspection of Table 2 reveals that overall direct experience did not




moderate cognitionndashbehaviour relations (high r+=51 vs low r+=45) Examining theimpact of direct experience as a moderator of attitudendashbehaviour intentionndashbehaviour and PBCndashbehaviour relations the following results were found participantswho received direct experience showed stronger attitudendashbehaviour relations (r+=43)than participants given indirect experience (r+=31) In addition participants whoreceived direct experience also demonstrated greater intentionndashbehaviour consistency(r+=67) than participants given indirect experience (r+=55) In contrast directexperience did not moderate PBCndashbehaviour relations (high r+=59 vs low r+=45)

InvolvementTable 2 shows that overall participants who were more involved with the attitudeobject showed greater consistency between attitude and intentionbehaviour (r+=57)than participants who were less involved (r+=31) In particular participants whowere more involved with the attitude object had greater attitudendashintention consistency(r+=58) than participants who were less involved (r+=31) In contrast participantswho were more involved did not possess stronger attitudendashbehaviour relations(r+=47) than participants who were less involved (r+=44)

CertaintyOverall participants with more certain cognitions showed a stronger relation betweencognitions and intentionbehaviour (r+=61) than participants with less certaincognitions (r+=25 see Table 2) Participants who were highly certain of their attitudesdemonstrated significantly stronger attitudendashbehaviour consistency (r+=46) comparedwith participants with low certainty in their attitudes (r+=08) Similarly the intentionndashbehaviour consistency of participants high in certainty was significantly greater(r+=64) than participants who were low in certainty (r+=41) Turning to cognitionndashintention relations participants high in certainty demonstrated significantly strongerattitudendashintention associations (r+=77) than participants low in certainty (r+=56) Inaddition participants with high certainty showed a larger subjective normndashintentionrelationship (r+=80) than participants with low certainty (r+=26)

AmbivalenceTable 2 summarizes the results for ambivalence Overall participants with low ambiva-lence possessed stronger associations between attitude and intentionbehaviour(r+=60) than participants with high ambivalence (r+=49) Meta-analysis demonstratedthat participants low in ambivalence possessed stronger attitudendashbehaviour con-sistency (r+=52) than participants high in ambivalence (r+=32) Similarly participantswith low ambivalence demonstrated stronger attitudendashintention relations (r+=65) thanparticipants with high ambivalence (high r+=56)

Affective-cognitive consistencyOverall moderation by affective-cognitive consistency was reliable (high r+=52 vslow r+=22 p lt 001) In addition participants with high affective-cognitive consist-ency demonstrated stronger attitudendashbehaviour consistency (r+=54) than participantswith low affective-cognitive consistency (r+=13) but affective-cognitive consistencydid not moderate the attitudendashintention relationship (both r+s=46)

Factors affecting moderation by propertiesSeveral factors are likely to affect how well properties moderate cognitionndashintentionand cognitionndashbehaviour relations The first ndash and most obvious ndash factor is the type of


behaviour under consideration Unfortunately studies of different moderator variableshave tended to examine different focal behaviours and there is no single behaviour thatpermits comparison of all seven moderators Moreover it was not possible to derive ameaningful system for categorizing behaviours that would allow us to determine theimpact of behaviour type on how well properties moderate particular relationshipsThus a limitation of the present review is that the effects obtained here must beinterpreted in the light of the fact that the moderator variables have been tested inrelation to different behaviours

However there were two factors that might affect moderation by properties ofcognitions that could be tested here namely publication status and the temporalcontiguity of cognition and behaviour measures Glass McGaw and Smith (1981)showed that publication status can influence effect sizes published studies generallyexhibit stronger effects compared with unpublished studies Sheeran and Orbell(1998) demonstrated that the time interval between measures of intention andbehaviour affected the strength of the intention behaviour relation correlations weresignificantly stronger over shorter than over longer time periods

Published vs unpublished studiesTo permit meaningful inferences we restricted the comparison of published versusunpublished studies to properties where there were at least two studies in both thepublished and unpublished categories Table 3 shows that accessibility and temporalstability were both reliable moderators of the intentionndashbehaviour relationship regard-less of publication status Accessibility moderated the attitudendashbehaviour relation inpublished studies and was associated with a marginally significant moderator effect inunpublished studies Z=151 p=066 Comparison of the moderator effect sizes forpublished versus published studies showed that there were no reliable differences inany of the three cases (all Zs lt 1) Thus publication status does not seem to constitutea substantial source of bias in the present review

Temporal contiguityTo examine the role of temporal contiguity in explaining moderator effects the timeinterval between measures of cognition and behaviour was coded in weeks for eachstudy (immediate post-tests were coded as zero weeks) We then compared the timeintervals employed in studies of different properties separately for attitudendashbehaviourand intentionndashbehaviour relations Findings indicated that time intervals were equiva-lent for moderators of the intentionndashbehaviour relation F(3 17)=096 ns Howeverthere was a reliable difference among time intervals for the seven moderators of theattitudendashbehaviour relation F(6 34)=435 p lt 05 Newman-Keuls post hoc testsindicated that the time interval between measures of attitude and behaviour wassignificantly longer in studies of temporal stability (M=22 weeks) than in studies ofaccessibility affective-cognitive consistency certainty and direct experience(Ms=170 100 100 and 040 respectively) None of the other pairwise comparisonswere reliable Thus temporal stability is probably disadvantaged compared to mostother moderators of the attitudendashbehaviour relation because studies of this propertyemployed longer time intervals

Comparisons of effect sizes for the moderator variablesAlthough the findings reported above suggest that moderator variables improvecognitionndashintention and cognitionndashbehaviour consistency it would be desirable to



compare the relative effect sizes of each moderator variable for each cognitionndashintention and cognitionndashbehaviour association to determine which are the most effec-tive moderators of these relationships To conduct this comparison the Z scoresobtained for each moderator variable were back-transformed into correlations andSchwarzerrsquos (1988) Meta program was used to test whether there were significantdifferences between the effect sizes for the moderators We first tested whether theeffect sizes for all the moderators of one relation (eg attitudendashbehaviour) were hetero-geneous If the x2 statistic was significant pairwise Z tests were conducted between allpairs of moderators to determine where were the differences between the variables

The x2 statistic was highly significant for the seven moderators of the attitudendashbehaviour relationship x2(6)=4007 p lt 001 Similarly the x2 statistic was significantfor moderators of intentionndashbehaviour consistency x2(3)=1240 p lt 001 Howeverthe effect sizes for the three moderators of PBCndashbehaviour relations did not differx2(2)=031 ns Finally there was significant heterogeneity among the moderators ofattitudendashintention associations x2(3)=1348 p lt 01 Thus pairwise Z tests were usedto compare the effect sizes reported for moderators of attitudendashbehaviour intentionndashbehaviour and attitudendashintention relations

Attitudendashbehaviour relationsAll seven variables included in this review have been employed as moderators of theattitudendashbehaviour relationship Table 4 shows the effect sizes for each moderatorvariable Temporal stability produced the largest effect size (r=37) and pairwiseZ tests demonstrated that temporal stability was a significantly better moderator ofattitudendashbehaviour consistency compared with all of the other moderator variables (allps lt 05) Certainty was the next strongest moderator variable and was a significantlybetter moderator of attitudendashbehaviour relations than accessibility ambivalence directexperience and involvement (all ps lt 05) Finally affective-cognitive consistency wasa more effective moderator of attitudendashbehaviour associations than were accessibilityambivalence direct experience and involvement (all ps lt 05) None of the othercomparisons yielded significant differences


Intentionndashbehaviour relationsAccessibility certainty direct experience and temporal stability have all beenemployed as moderators of intentionndashbehaviour consistency Table 4 summarizes theeffect sizes for these variables Temporal stability produced the largest effect size of thethree moderators (r+=25) and was a significantly better moderator of intentionndashbehaviour consistency than certainty (r=16 Z=181 p lt 05) and direct experience(r+=09 Z=318 p lt 001) and was a marginally better moderator than accessibility(r+=17 Z=148 p lt 07) However none of the other comparisons producedsignificant differences

Attitudendashintention relationsTable 4 shows that researchers have used affective-cognitive consistency ambivalencecertainty and involvement to moderate attitudendashintention consistency Certaintyproduced the largest effect size (r=19) and was a significantly better moderator ofattitudendashintention relations than both affective-cognitive consistency and ambivalence(both ps lt 05) Similarly involvement was a significantly better moderator of attitudendashintention relations than both affective-cognitive consistency and ambivalence (bothps lt 05) None of the other comparisons yielded significant differences

Discussion

This is the first quantitative review of the impact of moderator variables on cognitionndashintention and cognitionndashbehaviour relations Seven properties of cognitions wereexamined as potential moderators accessibility temporal stability direct experienceinvolvement certainty ambivalence and affective-cognitive consistency It waspredicted that participants who scored high on these variables (low for ambivalence)would exhibit significantly stronger cognitionndashintention and cognitionndashbehaviourrelations than participants who scored low on these variables (high for ambivalence)These predictions were largely confirmed and underline the utility of employingmoderator variables to enhance the prediction of intentions and behaviour

Accessibility was a reliable moderator of both attitudendashbehaviour and intentionndashbehaviour relations in the present review Moreover there was little evidence tosuggest that publication status affected the findings for this property Fazio (19891995 Fazio amp Towles-Schwen 1999) has claimed that accessibility is a highly import-ant aspect of attitude structure that may mediate the influence of other attitudinalproperties on attitudendashbehaviour relations Although the present findings confirmmoderation by accessibility results do not seem to be consistent with this mediationhypothesis accessibility produced one of the smallest effect sizes of the sevenattitudinal properties examined here and was a significantly weaker moderator ofattitudendashbehaviour relations compared with affective-cognitive consistency certaintyand temporal stability Thus it is unlikely that accessibility mediates moderationof attitudendashbehaviour consistency by affective-cognitive consistency certainty ortemporal stability (see Wegener and Fabrigar 2000 for a discussion of lsquomediatedmoderationrsquo) Further primary research is required in order to clarify whether otherattitudinal properties that had weaker moderator effects might be mediated byaccessibility

We found that direct experience moderated both attitudendashbehaviour relations andintentionndashbehaviour relations but did not moderate PBCndashbehaviour relations Thesefindings support Ajzen and Fishbeinrsquos (2001) claims that direct experience produces


more informative attitudes and intentions that are better predictors of subsequentbehaviour One reason why direct experience may moderate cognitionndashbehaviourrelations is that direct experience produces more stable cognitions (cf Doll amp Ajzen1992)

Two factors may explain the failure of direct experience to moderate PBCndashbehaviour relations in the two studies that have tested this relationship First Cookeand Sheeran (2004a) noted that their manipulation of direct experience (a single visitto wwwthehungersitecom) might not have been sufficiently strong to increase PBCndashbehaviour consistency among participants Second direct experience may not increasethe accuracy of PBC which is necessary for PBC to effectively predict behaviour(Ajzen 1991 Sheeran et al in press) Further research is needed to fully understandthe influence of direct experience on PBCndashbehaviour relations

Involvement moderated attitudendashintention but not attitudendashbehaviour relations Thefinding that involvement moderated attitudendashintention relations could be seen to sup-port Petty et alrsquos (1995) elaboration consistency hypothesis Petty et al argued that ahigh level of involvement leads to greater elaboration of information which in turnleads to attitudes that are more certain more accessible more internally consistentand therefore better predictors of intentionsbehaviour However it is important tonote that evidence for Petty et alrsquos account of how involvement affects intentions andbehaviour is mixed Lavine et al (2000) found that the impact of involvement onaccessibility was mediated by ambivalence and extremity Also Kokkinaki and Lunt(1997 1999) found that participants with high levels of involvement possessed moreaccessible attitudes compared with participants with low levels of involvement (seealso Bizer amp Krosnick 2001) However accessibility did not mediate the relationshipbetween involvement and behaviour as Petty et alrsquos hypothesis predicts MoreoverNederhof (1989) found no relation between involvement and certainty Althoughfurther primary research on involvement would be desirable some reformulationof the elaboration consistency hypothesis seems warranted in the light of availableevidence

Meta-analysis showed that certainty moderated attitudendashbehaviour intentionndashbehaviour attitudendashintention and subjective normndashintention associations One expla-nation of the moderating effects of certainty is that high levels of certainty couldproduce cognitions that are more stable and therefore better predictors of intentionsand behaviour Consistent with this hypothesis Bassili (1996) found that certainty wasthe best predictor of attitude stability in a study of nine attitudinal properties Bassilialso found that accessibility and certainty were independent predictors of stability (iecertainty was a significant predictor of stability even after accessibility had been takeninto account) indicating that the moderating effects of certainty were not mediated byaccessibility Sheeran and Abraham (2003) on the other hand found that the moderat-ing effect of certainty on intentionndashexercise behaviour relations was mediated by thetemporal stability of participantsrsquo intentions (see also Abraham amp Sheeran 2003) Thusevidence supports the view that certainty is an effective moderator of cognitionndashbehaviour relations because certainty is associated with greater temporal stability ofcognitions

Ambivalence moderated both attitudendashintention and attitudendashbehaviour relationsThere have been conflicting findings about the moderating role of ambivalence(Armitage amp Conner 2000 Jonas et al 1997) Some of this conflict is undoubtedlydue to the employment of different ambivalence measures (eg meta-judgmentalvs operative measures cf Conner amp Sparks 2002) The present review examined


operative measures of ambivalence only (ie studies that used either Thompsonet alrsquos 1995 formula or Kaplanrsquos 1972 formula) so it can be concluded thatambivalence is an effective moderator of attitudendashintention and attitudendashbehaviourrelations when this type of measure is employed Ambivalence may moderate attitudendashintention and attitudendashbehaviour relations because participants with more ambivalentattitudes also possess attitudes that are less stable Support for this idea comes fromboth survey research (Conner Sherlock amp Orbell 1998) and experimental studies(Bargh Chaiken Govender amp Pratto 1992) For example Bargh et al demonstratedthat participants who had less ambivalent attitudes gave more stable evaluations ofan attitude object at two time points compared with participants who had moreambivalent attitudes

We found that greater affective-cognitive consistency was associated with strongerattitudendashbehaviour correlations although affective-cognitive consistency did notmoderate attitudendashintention relations (note however that this latter finding was basedon just a single study) Affective-cognitive consistency may improve attitudendashbehaviourrelations because participants with high affective-cognitive consistency tend also topossess attitudes that are stable Indeed Rosenberg (1968) demonstrated that partici-pants with high affective-cognitive consistency possessed more stable attitudes 2weeks after the initial measurement of attitudes compared with participants with lowaffective-cognitive consistency (see also Erber et al 1995)

We expected temporal stability to moderate attitudendashbehaviour intentionndashbehaviour and PBCndashbehaviour relations These predictions were all supported bymeta-analyses showing that more stable cognitions provided better prediction ofbehaviour than less stable cognitions In fact the results for temporal stability showedthat this variable was the most effective moderator of attitudendashbehaviour relationshiphad the largest effect size for the intentionndashbehaviour association and was the onlyvariable that moderated PBCndashbehaviour relations There are two possible explanationsfor these findings First temporal stability may moderate cognitionndashbehaviour relationsbecause it predicts changes in cognition prior to action Second temporal stability maymoderate cognitionndashbehaviour consistency because it shields important cognitionsfrom competing cognitions

Ajzen and Fishbein (1973) claimed that certain conditions influence how effectiveintentions are at predicting behaviour ie when there is correspondence betweenmeasures of intentions and behaviour when behaviour is under complete volitionalcontrol and when intentions are measured in close temporal contiguity to measures ofbehaviour Further Ajzen and Fishbein argued that measuring intentions immediatelyprior to behavioural performance is critical because the longer the time intervalbetween measurement of intentions and behaviour the greater the likelihood thatparticipants will receive new information that may change their intentions Evidencesupports the view that shorter time intervals are associated with improved intentionndashbehaviour consistency (Sheeran amp Orbell 1998) The implication is that intentions thatare more stable provide a better indicator of the intention immediately prior to actionthan intentions that are less stable However if temporal stability is only importantbecause it is a marker for the intention immediately prior to action this implies thattemporal stability should moderate intentionndashbehaviour consistency over relativelyshort time periods only over longer time periods temporal stability should notmoderate intentionndashbehaviour relations because participants are likely to receivenew information that will alter the intention that is active immediately prior to thebehavioural performance


The results of a study by Conner Norman and Bell (2002) do not support Ajzen andFishbeinrsquos suggestion Conner et al found in the context of healthy eating behaviourthat participants with more stable intentions possessed stronger intentionndashbehaviourassociations compared with participants with less stable intentions even thoughbehaviour was measured 6 years after the measure of temporal stability had beentaken This finding is incompatible with Ajzen and Fishbeinrsquos prediction since theywould have expected that participantsrsquo intentions would change in 6 years because ofexposure to new information Thus the moderating effect of temporal stability is notexplained by its being a marker for intentions immediately prior to action as temporalstability can moderate intentionndashbehaviour relations even over extended periods oftime

An alternative explanation of moderation by temporal stability focuses on a mechan-ism by which temporal stability may moderate cognitionndashbehaviour relations It ispossible that temporal stability may represent the priority of intentions over time ieit provides an index of the extent to which an intention is shielded from competingintentions (cf Kuhl 1985) Kuhl suggested that successful completion of a goalrequires the protection of the intention to achieve the goal (eg lsquoI intend to givebloodrsquo) from other competing intentions that are unrelated to completion of the goal(eg lsquoI intend to go swimmingrsquo) Thus the temporal stability of a particular intentionmay reflect the position of that intention in a hierarchical structure of actions that theperson has told himherself to undertake in a particular time period (see also Karoly1998)

Consistent with the suggestion that temporal stability is an index of the priority ofparticular intentions is the idea that temporal stability could mediate the moderatingeffects of other variables Support for this hypothesis came from Sheeran and Abraham(2003) who tested the moderating effects of six variables (temporal stability certaintypast behaviour anticipated regret attitudinal vs normative control and self-schemas)on intentionndashbehaviour relations They found (a) that all six variables moderated theintentionndashbehaviour relationship (b) that each of the moderators was related totemporal stability and (c) that temporal stability mediated the impact of the otherfive variables ie when temporal stability was entered into the regression equationnone of the other moderator variables (interaction terms) were significant predictorsof intentionndashbehaviour consistency Research has also demonstrated that temporalstability can mediate the impact of direct experience on cognitionndashbehaviour rela-tions (Doll amp Ajzen 1992) These findings suggest that temporal stability may be themechanism through which other variables exert their influence and may be a criticalvariable in attempting to bridge the intentionndashbehaviour lsquogaprsquo (cf Sheeran 2002)

The hypothesis that temporal stability mediates the impact of other moderators hasnot been tested in the literature on attitudendashbehaviour consistency Indeed within theattitude strength literature temporal stability is viewed as a defining property of allstrong attitudes ndash together with resistance to persuasion impact on informationprocessing and behavioural impact (cf Krosnick amp Petty 1995) The possibility thattemporal stability might affect the behavioural impact of cognitions (as the presentanalyses demonstrated) does not appear to have been acknowledged However it isimportant to note as did Krosnick and Petty that little research has investigated therelations between these four characteristics and that it remains unclear whetherstrong attitudes possess all of these characteristics

The present analyses indicate that temporal stability and behavioural impact can bedistinguished on the dimension of antecedent versus consequent characteristics of


strong attitudes Whereas behavioural impact can only be conceptualized as a conse-quence of attitude strength temporal stability can be viewed either as a consequenceof attitude strength or antecedent to behavioural impact (Sheeran et al 1999)Relatively little research has been directed toward understanding the determinants oftemporal stability although indirect experience (Cooke amp Sheeran 2004a Doll ampAjzen 1992) ambivalence (Bargh et al 1992) and analysing the reasons underlyingattitudes (Wilson Dunn Kraft amp Lisle 1989) have all been associated with unstableattitudes Further research is warranted on the predictors of this property It is alsoworth pointing out that temporal stability could also be antecedent to the othercharacteristics of strong attitudes including resistance to persuasion and impact oninformation processing That is temporal stability might affect these characteristicssuch that more stable attitudes are more resistant to persuasion and have greatereffects on information processing compared with less stable attitudes Future studiesmight profitably be directed towards testing these possibilities in order to furtherincrease our understanding of temporal stability as a property of cognitions in thetheory of planned behaviour

In conclusion the present study demonstrated that seven properties ofcognitions ndash accessibility temporal stability direct experience involvement certaintyambivalence and affective-cognitive consistency ndash were all reliable moderators ofcognitionndashintention andor cognitionndashbehaviour relations Comparisons among thevariables indicated that temporal stability was the most effective moderator ofcognitionndashbehaviour relations However because studies of different moderatorsexamined different focal behaviours further (preferably experimental) research isneeded in order to examine the effects of these moderator variables in relation tothe same behaviour Such research would be valuable both to confirm the findingsobtained here and to test the hypothesis that temporal stability may mediate the effectsof the other moderators on cognitionndashbehaviour relations

Acknowledgements

We thank Mark Conner Paul Sparks and Greg Rennier for providing additionalinformation We also thank Geoff Haddock and three anonymous referees for their comments onan earlier version of this paper This research was funded by an ESRC award to the first author

References

An asterisk precedes references included in the meta-analysis

Abraham C amp Sheeran P (2003) Acting on intentions The role of anticipated regret BritishJournal of Social Psychology 42 495ndash511

Ajzen I (1985) From intentions to action A theory of planned behavior In J Kuhl amp JBeckmann (Eds) Action-control From cognition to behavior (pp 11ndash39) New YorkSpringer

Ajzen I (1991) The theory of planned behavior Organizational Behavior and HumanDecision Processes 50 179ndash211

Ajzen I (1996) The directive influence of attitudes on behavior In P M Gollwitzer amp J ABargh (Eds) Psychology of action (pp 385ndash403) New York Guilford

Ajzen I amp Fishbein M (1973) Attitudinal and normative variables as predictors of specificbehaviors Journal of Personality and Social Psychology 27 41ndash57

Ajzen I amp Fishbein M (2001) Attitudes and the attitude-behavior relation Reasoned andautomatic processes European Review of Social Psychology 11 1ndash33


Armitage C J amp Conner M (2000) Attitudinal ambivalence A test of three key hypothesesPersonality and Social Psychology Bulletin 26 1421ndash1432

Armitage C J amp Conner M (2001) Efficacy of the theory of planned behaviour A meta-analytic review British Journal of Social Psychology 40 471ndash499

Bargh J A Chaiken S Govender R amp Pratto F (1992) The generality of the automaticattitude activation effect Journal of Personality and Social Psychology 62 893ndash912

Baron R M amp Kenny D A (1986) The moderator-mediator variable distinction in socialpsychological research Conceptual strategic and statistical considerations Journal ofPersonality and Social Psychology 51 1173ndash1182

Bassili J N (1993) Response latency versus certainty as indexes of the strength of votingintentions in a CATI survey Public Opinion Quarterly 57 54ndash61

Bassili J N (1995) Response latency and the accessibility of voting intentions What contrib-utes to accessibility and how it affects vote choice Personality and Social PsychologyBulletin 21 686ndash695

Bassili J N (1996) Meta-judgmental versus operative indexes of psychological attributes Thecase of measures of attitude strength Journal of Personality and Social Psychology 71637ndash653

Berger I E amp Mitchell A A (1989) The effect of advertising on attitude accessibility attitudeconfidence and the attitude-behavior relationship Journal of Consumer Research 16269ndash279

Bizer G Y amp Krosnick J A (2001) Exploring the structure of strength-related attitudefeatures The relation between attitude importance and attitude accessibility Journal ofPersonality and Social Psychology 81 566ndash586

Budd R amp Spencer C P (1984) Latitude of rejection centrality and certainty Variablesaffecting the relationship between attitudes norms and behavioural intentions BritishJournal of Social Psychology 23 1ndash8

Conner M Norman P amp Bell R (2002) The theory of planned behavior and healthy eatingHealth Psychology 21 194ndash201

Conner M Povey R amp Sparks P (1995) The role of attitudinal ambivalence in the theoryof planned behaviour A study of low-fat diets Unpublished raw data School ofPsychology University of Leeds UK

Conner M Povey R Sparks P James R amp Shepherd R (2003) Moderating role ofattitudinal ambivalence within the theory of planned behaviour British Journal of SocialPsychology 42 75ndash94

Conner M Sheeran P Norman P amp Armitage C J (2000) Temporal stability as a moderatorof relationships in the theory of planned behaviour British Journal of Social Psychology39 469ndash493

Conner M Sherlock K amp Orbell S (1998) Psychosocial determinants of ecstasy use in youngpeople in the UK British Journal of Health Psychology 3 295ndash318

Conner M amp Sparks P (2002) Attitudes and ambivalence European Review of SocialPsychology 12 37ndash70

Conner M Sparks P Povey R James R Shepherd R amp Armitage C J (2002) Moderatoreffects of attitudinal ambivalence on attitudendashbehaviour relationships European Journal ofSocial Psychology 32 705ndash718

Cooke R amp Sheeran P (2001) Accessibility as a moderator of cognition-behaviourrelations Unpublished raw data University of Sheffield UK

Cooke R amp Sheeran P (2004a) Accessibility and temporal stability as moderators ofcognitionndashbehavior relations in the theory of planned behavior Manuscript submitted forpublication

Cooke R amp Sheeran P (2004b) The structure of properties of behavioural intentionand their capacity to moderate the intentionndashbehaviour relationship Manuscript underreview


Davidson A R amp Jaccard J J (1979) Variables that moderate the attitude-behavior relationResults of a longitudinal survey Journal of Personality and Social Psychology 37 1364ndash1376

Doll J amp Ajzen I (1992) Accessibility and stability of predictors in the theory of plannedbehavior Journal of Personality and Social Psychology 63 754ndash765

Erber M W Hodges S D amp Wilson T D (1995) Attitude strength attitude stability andthe effects of analyzing reasons In R E Petty amp J A Krosnick (Eds) Attitude strengthAntecedents and consequences (pp 433ndash454) Hillsdale NJ Erlbaum

Fazio R H (1989) On the power and functionality of attitudes The role of attitude accessibilityIn A R Pratkanis S J Breckler amp A G Greenwald (Eds) Attitude structure and function(pp 153ndash179) Hillsdale NJ Erlbaum

Fazio R H (1995) Attitudes as object-evaluation associations Determinants consequences andcorrelates of attitude accessibility In R E Petty amp J A Krosnick (Eds) Attitude strengthAntecedents and consequences (pp 247ndash282) Hillsdale NJ Erlbaum

Fazio R H Chen J McDonel E C amp Sherman S J (1982) Attitude-accessibility attitude-behavior consistency and the strength of the object-evaluation association Journal ofExperimental Social Psychology 18 339ndash357

Fazio R H Powell M C amp Williams C J (1989) The role of attitude accessibility in theattitude-to-behaviour process Journal of Consumer Research 16 280ndash288

Fazio R H amp Towles-Schwen T (1999) The MODE model of attitude-behavior processes In SChaiken amp Y Trope (Eds) Dual-process theories in social psychology (pp 97ndash116) NewYork Guilford

Fazio R H amp Williams C J (1986) Attitude accessibility as a moderator of the attitude-perception and attitude-behavior relations An investigation of the 1984 presidential electionJournal of Personality and Social Psychology 51 505ndash514

Fazio R H amp Zanna M P (1978a) On the predictive validity of attitudes The roles of directexperience and confidence Journal of Personality 46 228ndash243

Fazio R H amp Zanna M P (1978b) Attitudinal qualities relating to the strength ofthe attitude-behavior relationship Journal of Experimental Social Psychology 14398ndash408

Fisher R A (1921) On the probable error of a coefficient of correlation deduced from a smallsample Metron 1 3ndash32

Glass G V McGaw B amp Smith M L (1981) Meta-analysis in social research Beverley HillsCA Sage

Godin G amp Kok G (1996) The theory of planned behavior A review of its applications tohealth-related behaviours American Journal of Health Promotion 11 87ndash98

Godin G amp Conner M (2002) Temporal stability as a moderator of intentionndashbehaviourrelations Unpublished raw data University of Laval Canada

Godin G Conner M Gagnon H amp Lambert L-D (in press) Determinants of condom useamong a random sample of heterosexual adults British Journal of Health Psychology

Hedges L amp Olkin I (1985) Statistical methods for meta-analysis New York AcademicPress

Huiuk Y (1995) The influence of comparative advertising and amount of information onattitude accessibility Dissertations Abstracts International 56 (06) (University MicrofilmsNo AAC95-34266)

Hunter J E Schmidt F amp Jackson G B (1982) Meta-analysis Cumulating research findingsacross studies Beverly Hills CA Sage

Jonas K Diehl M amp Bromer P (1997) Effects of attitudinal ambivalence on informationprocessing and attitudendashintention consistency Journal of Experimental Social Psychology33 190ndash210

Kaplan K J (1972) On the ambivalence-indifference problem in attitude theory and measure-ment A suggested modification of the semantic differential technique PsychologicalBulletin 77 361ndash372


Karoly P (1998) Expanding the conceptual range of health self-regulation research Acommentary Psychology and Health 13 741ndash746

Kokkinaki F amp Lunt P (1997) The relationship between involvement attitude accessibilityand attitudendashbehaviour consistency British Journal of Social Psychology 36 497ndash509

Kokkinaki F amp Lunt P (1999) The effect of advertising message involvement on brandattitude accessibility Journal of Economic Psychology 20 41ndash51

Krosnick J A Boninger D S Chuang Y C Berent M K amp Carnot C G (1993) Attitudestrength One construct or many related constructs Journal of Personality and SocialPsychology 65 1132ndash1151

Krosnick J A amp Petty R E (1995) Attitude strength An overview In R E Petty amp J AKrosnick (Eds) Attitude strength Antecedents and consequences (pp 1ndash24) Hillsdale NJErlbaum

Kuhl J (1985) Volitional mediators of cognition-behavior consistency Self-regulatory processesand action versus state orientation In J Kuhl amp J Beckmann (Eds) Action control Fromcognition to behavior (pp 101ndash128) Heidelberg Springer

Lavine H Borgida E amp Sullivan J L (2000) On the relationship between attitude involvementand attitude accessibility Toward a cognitive-motivational model of political informationprocessing Political Psychology 21 81ndash106

Millar M G amp Millar K U (1996) The effects of direct and indirect experience on affectiveand cognitive responses and the attitude-behavior relation Journal of Experimental SocialPsychology 32 561ndash579

Millar M G amp Tesser A (1989) The effects of affective-cognitive consistency and thought onthe attitude-behavior relation Journal of Experimental Social Psychology 25 189ndash202

Nederhof A J (1989) Self-involvement intention certainty and attitudendashintention consistencyBritish Journal of Social Psychology 28 123ndash133

Norman R (1975) Affective-cognitive consistency attitudes conformity and behaviorJournal of Personality and Social Psychology 32 83ndash91

Petty R E amp Cacioppo J T (1986) Communication and persuasion Central and peripheralroutes to attitude change New York Springer

Petty R E Cacioppo J T amp Schumann D (1983) Central and peripheral routes to advertis-ing effectiveness The moderating role of involvement Journal of Consumer Research 10135ndash146

Petty R E Haugtvedt C P amp Smith S (1995) Elaboration as a determinant of attitudestrength Creating attitudes that are persistent resistant and predictive of behavior In R EPetty amp J A Krosnick (Eds) Attitude strength Antecedents and consequences (pp 93ndash130) Hillsdale NJ Erlbaum

Pieters R G M amp Verplanken B (1995) Intention-behaviour consistency Effects of consider-ation set size involvement and need for cognition European Journal of Social Psychology25 531ndash543

Prislin R (1996) Attitude stability and attitude strength European Journal of SocialPsychology 26 447ndash477

Regan D T amp Fazio R H (1977) On the consistency between attitudes and behaviorLook to the method of attitude formation Journal of Experimental Social Psychology 1328ndash45

Rennier G A (1989) The strength of the object-evaluation association the attitudendashbehaviourrelationship and the elaboration-likelihood model of persuasion Dissertations AbstractsInternational 50 (04) (University Microfilms No AAC89-15339)

Rosenberg M J (1960) A structural theory of attitude dynamics Public Opinion Quarterly 24319ndash341

Rosenberg M J (1968) Hedonism inauthenticity and other goals toward expansion of aconsistency theory In R P Abelson E Aronson W J McGuire T M Newcomb M JRosenberg amp P H Tannenbaum (Eds) Theories of cognitive consistency A sourcebook(pp 73ndash111) Chicago Rand McNally


Schwartz S H (1978) Temporal instability as a moderator of the attitudendashbehaviourrelationship Journal of Personality and Social Psychology 36 715ndash724

Schwarzer R (1988) Meta Programs for secondary data analysis Berlin Free University ofBerlin

Sheeran P (1999) Certainty of intention and intentionndashbehaviour relations Unpublishedraw data Department of Psychology University of Sheffield UK

Sheeran P (2002) Intention-behaviour relations A conceptual and empirical review EuropeanReview of Social Psychology 12 1ndash36

Sheeran P amp Abraham C (2003) Mediator of moderators Temporal stability of intentionand the intentionndashbehaviour relation Personality and Social Psychology Bulletin 29 205ndash215

Sheeran P amp Orbell S (1998) Do intentions predict condom use Meta-analysis andexamination of six moderator variables British Journal of Social Psychology 37 231ndash250

Sheeran P Orbell S amp Trafimow D (1999) Does the temporal stability of behavioralintentions moderate intention-behavior and past behavior-future behavior relationsPersonality and Social Psychology Bulletin 25 721ndash730

Sheeran P amp Taylor S (1999) Predicting intentions to use condoms Meta-analysis andcomparison of the theories of reasoned action and planned behavior Journal of AppliedSocial Psychology 29 1624ndash1675

Sheeran P Trafimow D amp Armitage C J (2003) Predicting behaviour from perceivedbehavioural control Tests of the accuracy assumption of the theory of planned behaviourBritish Journal of Social Psychology 42 393ndash410

Sparks P Conner M James R Shepherd R amp Povey R (2001) Ambivalence about health-related behaviours An exploration in the domain of food choice British Journal of HealthPsychology 6 53ndash68

Sparks P Hedderly D amp Shepherd R (1992) An investigation into the relationship betweenperceived control attitude variability and the consumption of two common foods EuropeanJournal of Social Psychology 22 55ndash71

Steffen V J amp Gruber V A (1991) Direct experience with a cancer self-exam Effects oncognitions and behaviour Journal of Social Psychology 131 165ndash177

Steffen V J Sternberg L Teegarden L A amp Shepherd K (1994) Practice and persuasiveframe Effects on beliefs intentions and performance of a cancer self-examination Journalof Applied Social Psychology 24 897ndash925

Thompson M Zanna M amp Griffin D (1995) Letrsquos not be indifferent about (attitudinal)ambivalence In R E Petty amp J A Krosnick (Eds) Attitude strength Antecedents andconsequences (pp 361ndash386) Hillsdale NJ Erlbaum

Trafimow D (1994) Predicting intentions to use a condom from perceptions of normativepressure and confidence in those perceptions Journal of Applied Social Psychology 242151ndash2163

Trafimow D amp Sheeran P (1998) Some tests of the distinction between cognitive andaffective beliefs Journal of Experimental Social Psychology 34 378ndash397

Trafimow D Sheeran P Conner M amp Finlay K A (2002) Is perceived behavioural control amultidimensional construct Perceived difficulty and perceived control British Journal ofSocial Psychology 41 101ndash121

Verplanken B (1989) Involvement and need for cognition as moderators of beliefs-attitudendashintention consistency British Journal of Social Psychology 28 115ndash122

Warland R H amp Sample J (1973) Response certainty as a moderator variable in attitudemeasurement Rural Sociology 38 174ndash186

Wegener D T amp Fabrigar L R (2000) Analysis and design for nonexperimental dataAddressing causal and noncausal hypotheses In H T Reis amp C M Judd (Eds) Handbookof research methods in social and personality psychology (pp 412ndash450) CambridgeCambridge University Press


Wilson T D Dunn D S Kraft D amp Lisle D J (1989) Introspection attitude change andattitude-behavior consistency The disruptive effects of explaining why we feel the way wedo In L Berkowitz (Ed) Advances in experimental social psychology (Vol 22 pp 249ndash263) San Diego CA Academic

Winter L amp Goldy A S (1993) Effects of prebehavioural cognitive work on adolescentsrsquoacceptance of condoms Health Psychology 12 308ndash312

Received 8 January 2002 revised version received 29 November 2002


the personrsquos actual control over behavioural performance (Sheeran Trafimow ampArmitage 2003)

Meta-analyses indicate that the TPB provides a good explanation of a wide range ofbehaviours For example attitudes subjective norms and PBC account for 39ndash42 ofthe variance in intentions (Armitage amp Conner 2001 Godin amp Kok 1996 Sheeran ampTaylor 1999) while intentions and PBC predict between 28 and 34 of the variance inbehaviour (Armitage amp Conner 2001 Godin amp Kok 1996 Trafimow Sheeran Conneramp Finlay 2002) Notwithstanding the success of the TPB there remains a substantialproportion of variance in intentionsbehaviour that is not explained by TPB variables(Sheeran 2002)

One line of research designed to increase the predictive validity of the TPB has beento examine variables that might moderate cognitionndashintention and cognitionndashbehaviour relations A moderator variable affects the relationship between two othervariables (cf Baron amp Kenny 1986) For example the relationship between exerciseattitudes and exercise behaviour may be moderated by the certainty of participantsrsquoattitudes such that participants who possess highly certain attitudes exhibitstronger attitudendashbehaviour relations than participants who are less certain of theirattitudes The present study focuses on seven properties classified by Krosnick andPetty (1995) as relating to defining features of attitude strength (temporal stability)aspects of attitude structure (accessibility affective-cognitive consistency ambiva-lence) subjective beliefs about attitudes (certainty involvement) or processes ofattitude formation (direct experience)

Properties of cognitions as moderator variables

Krosnick and Petty (1995) suggested that all seven properties are likely to improveattitudendashbehaviour consistency though they also note that these properties are oftencorrelated with one another This could indicate that these properties improveattitudendashbehaviour consistency because they reflect a single underlying constructHowever contrary to this hypothesis research by Krosnick Boninger Chuang Berentand Carnot (1993) suggested that properties of attitudes are conceptually distinctUsing confirmatory factor analysis Krosnick et al (1993) found that a single-factormodel provided an inadequate description of the structure of ten attitudinal propertiesdespite significant correlations between the properties Exploratory factor analysesfound the best fit for a five-factor model with no more than three properties loadingon any one factor (see also Bassili 1996 Erber Hodges amp Wilson 1995 Prislin 1996)Thus although properties of attitudes may be correlated with one another they areconceptually distinct constructs

Researchers have also employed these properties as moderators of relationswithin the TPB Studies have demonstrated that attitude strength indices canmoderate attitudendashintention relations (Armitage amp Conner 2000 Budd amp Spencer1984) and can be adapted to moderate subjective normndashintention (Nederhof 1989Trafimow 1994) intentionndashbehaviour (Bassili 1995 Sheeran Orbell amp Trafimow1999) and PBCndashbehaviour (Conner Sheeran Norman amp Armitage 2000 Doll ampAjzen 1992) relations The aim of the present study is to conduct the first quantitativereview of the impact of the seven moderator variables outlined above on cognitionndashintention andor cognitionndashbehaviour relations In the next section we define eachof the properties and describe previous research that tested moderation by theseproperties
















The present study


Method














Results































Discussion
























Acknowledgements


References
















































































































The present study


Method














Results































Discussion
























Acknowledgements


References











































































































Results































Discussion
























Acknowledgements


References






























































































































Discussion
























Acknowledgements


References





















































































































Acknowledgements


References
































































































