Professional Documents
Culture Documents
816 e 3462 FBD 9 Aef 8 A 550
816 e 3462 FBD 9 Aef 8 A 550
University of Florida
Paul W. Miniard is now a member of the Faculty of Marketing, Ohio State University.
Parts of this research are based upon the first author’s masters thesis prepared for the
Department of Marketing, University of Florida, Gainesville. Appreciation is expressed to
Olli T. Ahtola, Stephen A. LaTour, and Barry R. Schlenker who served as committee
members, to James L. Ginter for his analysis of multicollinearity issues, and to Deborah
Miniard for her assistance on the project. We also wish to acknowledge the support provided
by the Center for Consumer Research at the University of Florida.
Requests for reprints should be sent to Joel B. Cohen. Center for Consumer Research,
University of Florida. Gainesville, FL 3261 I.
309
0022-1031/81/030309-31$0200/O
Copyright @ 1981 by Academic Press. Inc.
All rights of reproduction in any form reserved.
310 MINIARDANDCOHEN
The first component, A,, is the actor’s attitude toward performing the
behavior in question under a given set of circumstances. In accordance
with an expectancy-value formulation, a person’s attitude toward a spe-
cific behavior is proposed to be a function of the salient beliefs about the
relevant attributes and perceived consequences of performing the
behavior and of the person’s evaluation of these attributes and conse-
quences:
A, = pbiei, (2)
i=l
SN = 2 NBiMCi, (3)
i=l
where NB stands for normative belief (i.e., the person’s belief that refer-
ence group or individual i thinks he should or should not perform the
behavior); MC is the person’s general motivation to comply with referent
i: and n is the number of relevant referents.
The relative importance of these two components in determining inten-
tions is expected to vary with the behavior, with the situation, and with
individual differences among actors. Component weights (i.e., u’, and ~1~)
for As and SN are traditionally estimated by multiple-regression proce-
dures and are interpreted as being “proportional to their relative impor-
tance in the prediction of behavioral intentions” (Fishbein & Ajzen, 1975.
pp. 302, 303).
According to Fishbein and Ajzen (1975, pp. 368-372). the relationship
between behavior and behavioral intentions depends upon (1) the degree
to which the measure of intention corresponds directly to the observed
behavior, (2) the time interval between measurement of intention and the
behavior’s occurrence, and (3) the degree to which the actor is able to act
in accordance with his or her own intentions or without the assistance of
others. A stronger relationship between intention and behavior should
occur when the intention measure is specific to the behavior of interest,
performance of the behavior is temporally close to the measurement of
intention, and the behavior is under volitional control.
Separating Attitudinal and Normative Influences
In modifying Dulany’s (1961. 1968) theory of propositional control
dealing largely with verbal conditioning, the model’s components were
312 MINIARD AND COHEN
’ ConverseI). predictors that are not related to the criterion may appear important as
reflected by a significant beta coefficient. Such occurrences have been discussed under the
label of “suppressor” variables (cf. Cohen & Cohen, 197.5).
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 313
Statements of this kind suggest that the model’s components are intended
to provide a distinct representation of these two sources of influence.
Fishbein and Ajzen’s basis for distinguishing attitudinal and normative
influences on behavioral intentions is explicit though largely operational:
“The theory suggests that it is useful to maintain the distinction between
beliefs about the consequences of perj&wzing a behavior and beliefs about
expectations of relevant rejkrents.” Thus, using an example from Fish-
bein and Ajzen (p. 304). a belief that “My child thinks I should buy Sugar
Puffs” is regarded as normative on the basis that it refers specifically to a
referent’s expectation regarding the behavior in question. On the other
hand, a belief that “Buying Sugar Puffs will please my child” would not
be regarded as normative since this belief, as stated, does not contain an
explicit referent expectation.
Fishbein and Ajzen have offered no evidence that people, in fact,
maintain such a distinction in their thinking. So, once we move beyond a
surface level (both NB and SN measures have been designed to separate
out the “X thinks I should” type of belief), we are still faced with the
problem of establishing that these are distinct in any important psy-
chological sense. Thus, the belief, “Buying Sugar Puffs will please my
child” and the belief “My child thinks I should buy Sugar Puffs.” while
structurally different, may reflect a similar underlying concern with the
child’s reactions. The second belief might result from a direct statement
by the child, and the first might very well be a direct implication of that
statement. For this reason it is not clear why so much should be made of
this surface distinction. What may be much more significant, however, is
whether or not these two beliefs reflect a common underlying concern
with pleasing the childfijr normati~~e reasons (e.g.. to avoid a temper
tantrum or have the child respond with praise).
It is our contention that the conceptual basis underlying Fishbein and
Ajzen’s separation of attitudinal from normative influences is also in-
adequate if these are to be interpreted as separate sources of influence on
behavioral intentions. To illustrate this, let’s assume that the true state of
the world is as follows. You essentially have no attitude toward buying
314 MINIARD AND COHEN
Moreover, this distinction emphasizes the importance of two basic social psy-
chological concepts that have traditionally been treated independently. Psycholo-
gists and sociologists interested in individual behavior have frequently made use of
the attitude concept whereas theorists dealing with groups and societies have often
relied on the concept of social norms. By including an attitudinal and a normative
component. the present theory emphasizes the importance of both concepts and
provides a bridge between the two approaches to the study of human behavior”
(Fishbein & Ajzen. 1975, p, 304, 305).
316 MINIARDANDCOHEN
If, in fact. a major goal of the model is to represent and separate “these
basic social psychological concepts.” this approach seems conceptually
flawed. The traditional distinction between complying with others for
normative reasons and engaging in a behavior because it is consistent with
one‘s attitudes and values is not only lost, it is made trivial. Thus. to take
a further example from Ajzen and Fishbein (l%O), the belief “my hus-
band thinks I should have an abortion” is regarded as normative whereas
the belief “my having an abortion would please my husband” would be
attitudinal (p. 73).
The lack of a clear separation can be seen in the measures. In measuring
normative beliefs (NB). the respondent is asked to indicate whether
“Referent X thinks I should/should not perform behavior Y.” Thus. I
may believe that my attorney thinks I should sign a contract. but such
expert information should also lead to the creation or change of beliefs
(e.g.. that the contract is to my advantage, protects my rights) that
determine my personal attitude toward the behavior.
Similar problems appear inherent in the measurement of motivation to
comply (MC) which asks the respondent how much “I want to do/I want
to do the opposite of what referent X thinks I should do.” Since this
operationalization does not restrict itself to compliance but extends to any
reason for wanting to engage in a behavior you believe the referent
endorses for you. it should incorporate aspects of personal as well as
normative reasons for engaging in a behavior.
In the most recent approach to measuring the normative component.
subjective norm (SN) is operationalized as “Most people who are impor-
tant to me think 1 should/should not perform behavior Y.” The measure
does not distinguish between others who are important for informational
as opposed to normative reasons. In addition, as elaborated in a sub-
sequent section. this me;;sure simply assesses one’s perception of what
important others think the person should do. Yet. SN is thought to
incorporate both specific normative beliefs and motivation to comply with
them. There is a further ambiguity inherent in an SN-type measure since
respondents may define the “behavior” so as to include complying (01
not) with influence pressures in the immediate situation (e.g.. voting
“yes” is coded as voting “yes” to comply with a manipulator). Since
people probably feel that others would not want them to comply with a
manipulator. such a recoding of the behavior should make the SN mea-
sure ambiguous.
Turning to the attitudinal component, measures of AN have typically
asked subjects to simply evaluate their “performing behavior X.” Since
subjects are not asked for their independent evaluation (i.e., how they feel
about the behavior unmindful of normative outcomes). it is likely that
subjects will. to some extent. include normative considerations in their
responses.
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 317
Research to date has indicated relatively little variance in this measure. and thus
the results obtained with normative beliefs alone were as good or better than those
obtained when NB was multiplied by MC (p. 469).
Fishbein (1976) has stated that scoring his previous unipolar MC mea-
sure (i.e., the scale ranged from “I want to do” to “I want not to do”) as a
bipolar scale was “probably inappropriate” and that “as a next step we
should measure motivation to comply on a true bipolar scale” (p. 494).
Such bipolar concepts would be necessary to represent social “influ-
ence” that one would not simply disregard but instead would lead one to
behave in an opposite manner (e.g., reactance theory predictions, rebel-
lion against authority). Therefore. if the psychologically relevant end-
points are “1 want to do” and “I want to do the opposite of,” these
should be assessed using a bipolar scale. This alteration in scale format
may not by itself lead to findings supporting the predictive utility of MC.
Research (see, for example, Bass & Wilkie, 1973; Cohen & Ahtola. 1971;
Wilkie & Pessemier. 1973) addressing this issue with respect to the at-
titudinal component (i.e.. the utility of weighting beliefs by values) has
cautioned researchers not to expect increased power where there is low
variability in the evaluative measures (i.e.. when values assigned to at-
tributes are relatively equal). Similarly. MC should contribute little to the
prediction of BI, regardless of how the scale is coded. when positive
referents are of roughly equal importance (i.e., when MC varies little
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 319
Accurate prediction of a given intention. however. can be expected only when the
attitudinal and normative components of the model are measured at the same level
of specificity as is the intention (Fishbein & Ajzen. 1975, p. 333).
A total of 193 male and female students. enrolled in various undergraduate marketing
courses at the University of Florida, participated in the experiment. After reading a scenario
in which a hypothetical woman was considering the purchase of a dress, subjects were asked
to play the role of this woman in responding to a questionnaire containing the measures
necessary for operationalizing the model. Subjects were randomly assigned to the cells of a
2 x 2 x 3 factorial design. The first factor (attitude toward object) was designed to
manipulate the attitudinal component by varying the woman’s attitude toward the dress, the
second (referent inguence potential) the normative component through the presence or
absence of information that the woman’s friend was exerting influence for self-serving
reasons, and the third (MC spec$cit~) involved the measurement of MC at three alternative
levels of specificity. These separate manipulations of attitudinal and normative influence
allow an examination of the degree to which each component captures a separate set of
influences on behavioral intentions.
manipulation was also intended to create a situation in which one is not motivated to comply
with a generally influential referent, thus permitting a test of the adequacy of measuring MC
at a general level.
Before proceeding to the questionnaire, subjects were asked to return to the scenario and
circle those aspects of the situation which they considered important. Then. after being
reminded to assume the role of Debbie. they responded to one of three questionnaires (the
third between-subjects factor in the design) which differed only in the level of specificity at
which MC was assessed (i.e., general. moderate. and situation specific). This factor was
included to examine the implications of measuring MC at different levels of specificity.
Deprrldrr~r Vuriahles
Since prior research varying item order showed no significant differences (Ajzen &
Fishbein. 1969). we employed the order suggested by the schematic representation of the
model as presented by Fishbein and Ajzen ( 1975, p, 334). Therefore. subjects first responded
to one of three alternative measures of MC via 7-point scales ranging from “I want to do”
(+3) to “I want to do the opposite of” (-3). Depending upon whether level of specificity
was general, moderate, or situation specific, the scale was prefaced by either “In general.”
“In the area of clothes shopping behavior,” or “In this situation.” respectively. Next. NB
was assessed on 7-point scales with the endpoints *‘I should” (+3) and “I should not” (-3).
NB and MC were obtained for three referents: the hypothetical person’s neighbor (Sue).
husband. and parents.’
Subjects then indicated their attitude toward the behavior (AR). They received Fishbein‘s
standard (Ajzen & Fishbein, 1969, 1970. 1972) lead-in (“Buying the dress is:“) and then four
7-point bipolar evaluative semantic differential scales (good-bad. foolish-wise.
rewarding-punishing. harmful-beneficial). Subjects next responded to theSN measure via a
7-point scale ranging from “I should” (+3) to “I should not” (-3). This was followed by an
alternative attitudinal measure (PA,), not employed by Fishbein. This measure (“Forgetting
what others think you should do and their reaction. youpersonally feel that buying the dress
is:“) was designed to minimize normative influences by instructing subjects that the influ-
ence of others was to be ignored. The major difference. then. between PA, and A, was that
the latter does not suggest that the influence of others is to be ignored.
BI was next measured on three 7-point bipolar semantic differential scales (likely-
unlikely, probable-improbable. possible-impossible). Perceptions that the referent was
trying to help or manipulate the person were also separately assessed on these same three
semantic differential scales and combined into an index so as to provide a check on the
referent injhencc~ potential manipulation.
2 Three referents were used because NB for Sue was held constant across the scenarios
(i.e.. the state of disagreement was manipulated by changes in attitude and not by variations
in the referent’s expectations). The lack of variance in NB would constrain correlations with
SN. so that even a “weak” measure of MC could improve this fit. To avoid this statistical
problem in evaluating SN = I; NBMC predictions, additional variance was built in by using
the husband and parents as additional referents across all scenarios. The choice of these two
referents was based on the fact that the hypothetical person’s husband was mentioned in
each of the scenarios. and both husbands and parents might well be relevant referents when
the “wisdom” of complying with a new acquaintance would be an issue. In analyses looking
instead at the sensitivity of a particular measure to alternative sources of influences that do
not involve correlation methods. only the one referent specifically used to create variance in
normative influence was employed. Unless specified in the text. NB and MC scores for all
three referents (i.e.. a NBMC score is computed for each referent, summed across the three
referents, and the resultant score correlated with the criterion ofinterest) will only be used in
testing correlational hypothesis (i.e., model predictions).
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 323
Subjects were next asked how difficult it was to play the role on a S-point scale ranging
from ‘not at all difficult” (I) to “extremely difficult” (5) and how confident or certain they
were that their responses would be similar to the hypothetical person’s responses on a
6-point scale ranging from “extremely certain” (I) to “extremely uncertain” (6). These were
included to provide a check both on any particular problems subjects had playing the role in
these scenarios and for potential differences between male and female abilities to capture the
role.
Supplementary Measures
Measures addressing the validity of the scenario manipulations were collected from two
groups of subjects who did not take part in the main experiment. Each subject randomly
received one of the scenarios. After reading the scenario, the first set of subjects (II = 67)
were asked to evaluate nine social outcomes (as assessed by scales measuring the value of
being liked and disliked by, pleasing and displeasing, gaining the approval and disapproval
of, being similar and dissimilar to, and obtaining a favorable reaction from the referent)
potentially mediated by Sue (the referent) on 7-point scales ranging from “extremely good”
(+3) to “extremely bad” (-3) in order to show that the referent’s influence potential was in
fact reduced by displaying her manipulative intent. The referent’s influence potential should
be reduced by the inclusion of information describing the referent’s manipulative intent. but
not by the attitudinal manipulation.
The second group (n = 57) was given a “referent influence” scale and was asked to
estimate the amount and direction of social influence existing in the situation via an I l-point
scale with the endpoints “more likely to buy the dress” (1 I) and “less likely to buy the
dress” (I). The scale midpoint was to be marked if the referent’s influence was unimportant.
The goal of this check was to evaluate the impact of the refcwnf influence porentiul
manipulation on perceptions of referent influence.
RESULTS
Manipulation Checks
Three separate manipulation checks were used to evaluate the ade-
quacy of the referent influence potential treatment. The first examined the
overall impact of this particular manipulative behavior on others’ evalua-
tion of outcomes mediated by that referent. The second assessed that
subjects did in fact perceive the referent as behaving in a manipulative
fashion, and the third examined the perceived influence of this manipula-
tive behavior on the likelihood of subjects’ buying the dress.
Social outcome measures. Each of the nine ‘-social outcome” mea-
sures was tested individually using a 2 (attitude toward object) x 2
(referent infkence potential) analysis of variance.3 The latter factor was
significant @ < .OOl) in each case and accounted for, on average, 29% of
:% Cell sizes ranged from I3 to 18. As a result of the unequal and disproportional cell sizes,
unless particular procedures are undertaken, an orthogonal partitioning of the explained
variability is not achieved, and the resulting tests of significance are biased (Applebaum &
Cramer. 1974). Therefore, for this and subsequent ANOVAs, an analytical procedure known
as experimental design analysis (Overall & Speigel. 1969) was employed. Basically, this
analysis involves the estimation of a main effect by adjusting for the remaining main effects
but ignoring interactions. and then estimation of interactions through adjusting for both main
effects and interactions of the same and lower order.
324 MINIARD AND COHEN
the measures’ variability using the & statistic (Hays, 1967). thus provid-
ing support for this manipulation’s impact on factors assumed to affect
one’s motivation to comply with a referent. There were no significant (p >
.05) effects of the attitudinal manipulation on any of the nine measures.
The latter analysis was carried out to determine if disagreement between
the person and the referent as to whether or not the person should buy the
dress would lower perceived referent attractiveness and hence the refer-
ent’s influence potential. If such an effect occurred. the “confounded”
attitudinal manipulation would make an unambiguous assessment of the
normative measures’ attitudinal contamination impossible.
Prrcei\-cd rmnipulcrti~~e intent. To verify that the referent was per-
ceived to be manipulative, the 193 experimental subjects’ perceptions of
the referent’s manipulative intent were analyzed as a function of attitude
tmwrd object, reJ&~tzt infirrencr pote~rticrl, and the level of specificity of
the motivation to comply measure (i.e., the questionnaire manipulation).
As expected, only referent injrrrnw potential was significant [ F( 1, IX I ) =:
949.31, p < .OOl] and accounted for 83% of the variation.
Ryfcrerlt injlucnce s~ult~. Finally. both r&rent irljhrencr potentiul con-
ditions pushed subjects’ responses away from the scale midpoint (i.e., 6).
Cell means for the positi\-c and nrgrrtil,e referent i@uence potential
scenarios were 8.31 and 4.14. respectively. and both were significantly
different from the scale midpoint. The ctttitctdc> to\cm-d ohjrct treatment
means were also tested against this criterion. Neither of the means sig-
nificantly (P > .05) departed from the scale midpoint. again supporting
this factor‘s lack of impact on the referent’s influence potential.
Behavioral Intrntions
Before examining the measures’ sensitivity to the experimental manipu-
lations. we tested whether our manipulations significantly altered be-
havioral intentions (Bl). If intentions were not affected by such manipula-
tions. it might indicate that only a minimal impact on the determinants of
intentions was achieved by these manipulations. Using a 7 x 2 v 3
ANOVA. both the crttitudr tmard object [F( I. 181) = 116.72, p < .OOl]
and the reJ>rrnt inj2nenc.epotential [F( 1. 181) = 61.03, p < .OOl] factors
were highly significant and explained 31% and 16%’of the variability in BI.
respectively. A significant [F( I. 181) = 4.75. p < .03 I, A2 = ,011 interac-
tion was also detected, indicating that the difference in BI between the
attitude conditions was greater when the referent was not seen as a
manipulator than when she was.
The reasonableness of employing these data to evaluate the measures
was further supported since the model provided an acceptable prediction
of Bf. The results of pooled (i.e., collapsing across experimental cells
after adjusting for mean differences between cells) multiple regression
analyses for models containing the various attitudinal and normative
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 325
measures are summarized in Table 1.4 These data give an early indication
of the weaknesses of the more general MC measure as the normative
component’s partial regression coefficient failed to achieve significance
when MC was operationalized at this global level.”
TABI,E I
Poo~m MULJIPLE KEGRESSI~N ANALYSES POR SHE. PREDICTION OF
BEHAVIORAL INTENTIONS
Model II B;’ B$ R R’ P’
NOIP. The symbols MCg, MCnz. and MCs refer to whether motivation to comply was
measured at a general. moderate. or situation-specific level, respectively.
‘I Standardized partial regression coefficient for attitudinal component.
b Standardized partial regression coefficient for normative component.
c Pooled simple correlation (i.e., multicollinearity) between attitudinal and normative
components.
* p i .os.
** p < .Ol.
factors.i The cell means for these measures as a function of the normative
and attitudinal manipulations and the F ratios for these two main effects
and their interaction are presented in Table 2. The MC specijicity factor
was excluded from the table since it failed to achieve significance as either
a main effect or interaction in any of these analyses.
If MC taps distinguishable normative influence-based responses (i.e., if
it is not also sensitive to attitudinal influence), subjects’ responses to the
measure should be affected only by the r&rent ir~Jiurncr potentid ma-
nipulation. While this factor reached significance (& = .17), the attitudinal
manipulation had a smaller but significant (& = .07) impact on the measure.
As suggested earlier, therefore, MC does not appear to isolate compliance-
based reasons for wanting to engage in a behavior from attitudinal ones.
Differences in the attitudinal manipulation’s impact upon MC as a
function of the measure’s level of specificity were not found as the
interaction between the questionnaire and attitudinal factors failed to
approach significance @ > . I). Similarly, the notion that a situation-
specific measure would be superior to more general measures of MC in
capturing social influence was not supported, as the interaction between
the questionnaire and normative manipulations was also insignificant.
Possible explanations for these nonsignificant interactions are presented
in the Discussion.
i There were no significant effects of the manipulations beyond those specihcally consid-
ered in the discussion of study goals.
TABLE 2
MEANS AND F RATIOS FOR NORMATIVE AND A~rrmumNAL MEASURES RY SCENARIO MANIPULATIONS
F Ratios
Referent influence potential
Referent influence Attitude toward
Measure/attitude toward object Positive Negative potential object Interaction
Nore. The normative measures had a scale range of +3 to -3 whereas the attitudinal measures ranged from + I2 to - 12.
* p < .05.
** p < .OOl.
328 MINIARD AND COHEN
” As the manipulation of social influence was achieved by varying the referent’s influence
potential and not through changes in perceived referent expectations. it was not possible to
test NB’s responsiveness to normative factors. Despite the conceptual similarity ofNB and
SN. SN’s sensitivity to the normative influence manipulation was tested since Fishbein
(1976. p. 496; Fishbein & Ajzen. 1975. Chap. 7) seems to be recommending the use of SN as
a replacement for ZNBMC. It would appear that SN is expected to reflect any changes in
normative influences. regardless of their perceptual or motivational basis.
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 329
TABLE 3
POOLED SIMPLE CORRELATIONS BETWEEN BEHAVIORAL INTENTIONS AND
NORMATIVE MEASURES
TABLE 4
POOLED SIMPLE CORRELAIXON BETWEEN SUBJECTIVE NORM AND
ITS HYPOTHESIZED DETERMINANTS
(’ Column values represent the level of significance for the predictive difference between
CNB and XNBMC.
* Row correlations are pooled across all experimental factors.
* ,> < .05.
** ,’ < .Ol.
!’ Fishbein and Ajzen question our conclusions concernmg the inadequacy of assessing
MC at a general level and SN’s inability to mediate the MC-B1 relationship. They argue that
(/ we coded our bipolar ‘WC measure in a unipolar fashion and if we had computed
unadjusted correlations. we would have reached different conclusions. They note in their
paper that they had originally coded their unipolar motivation to comply scale (“want
to-want not to.‘) in a bipolar fashion: however, when they subsequently restored this scale
in keeping with its unipolar format. the correlations involving MC increased. Concerning the
first point, Fishbein t 1976) earlier argued for the use of a true bipolar MC scale. This is the
scale we used and, of course. scored in keeping with the responses subjects thought they
were supplying (i.e.. bipolar scoring). Fishbein and Ajzen now argue that we should score
motivation to comply in a unipolar fashion.
But assigning a negative referent a positive motivational score seems to us both atheoreti-
cal and counterintuitive. Moreover. the improvement in correlations may merely reflect a
scale transformation artifact that results when a positive constant is used to eliminate
negative scale values and there is a preponderance of positive referents in the data base.
Given a majority of negative referents in the data base. the opposite scale transformation
(i.e., from unipolar to bipolar) could artifically increase correlations. Given characteristics of
the data base, improvement in correlations can always be achieved through a specified scale
transformation. We simply believe that transforming our bipolar scale to a unipolar scale
and then using these data in correlational analyses is unsound.
As discussed in footnote 4, the appropriateness of pooled versus unadjusted correlations
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 333
is not an easily resolved issue. We therefore reanalyzed our data without adjusting for mean
differences and found that the major conclusions bearing on MC‘s predictive value and the
SN = ZNBMC relationship are unaffected when the motivation to comply scale is scored in
its original bipolar fashion.
334 MINIARD AND COHEN
by the results of our experiment, the MC and SN measures are more likely
to be influenced by manipulations of one’s attitude than is NB. Since these
measures were not included in their study, the question of whether a
manipulation can be developed that impacts upon only the A, measure
and not the MC and SN measures is not resolved by this study.“’
In defense of their distinction between the attitudinal and normative
components, Fishbein and Ajzen present several examples which they
feel demonstrate the lack of redundancy among behavioral and normative
beliefs and that this is more than a surface distinction. One of these
examples deals with a situation in which a person believes that buying his
wife a diamond ring would make her happy (in their system this would
lead to a positive attitude). At the same time the husband believes that his
wife thinks he should not buy the ring (this would lead to a negative
subjective norm). This example is obviously incomplete: his wife’s overall
negativity toward his buying the ring stands in marked contrast to his
belief that such behavior would make her happy. Fishbein and Ajzen
parenthetically suggest that possibly the wife feels they cannot afford the
ring. It seems more parsimonious to say that the wife’s summary judg-
ment is negative and that, assuming one wanted to understand this
further, this judgment is based on her more heavily weighted belief that
they can’t afford the ring and the less heavily weighted belief that she
would enjoy wearing it. Assuming that one wanted to understand the
factors impacting on the husband’s behavioral intention, we feel it would
be better to differentiate clearly between the husband’s desire to make his
wife happy out of a sense of gratitude or love (a personal reason) as
opposed to wanting to evoke favorable reactions from her (a normative
reason). Our approach would suggest. then, that the key distinction is not
between a behavioral belief and a referent expectation but between per-
sonal and normative reasons for engaging in a behavior.
In their paper, Fishbein and Ajzen note that the original presentation of
our study design explicitly questioned the construct validity of their
‘” Fishbein and Ajzen contend that the motivational components of their model (i.e.. MC
and e,) should be assessed “independent of the behavior in question.” In part they justify
this by adding that “Neither evaluation nor motivation to comply are used as direct
predictors of intention: they merely modify the behavioral or normative beliefs which are
measured in correspondence with the intention and the behavior.” We do not feel this is a
very persuasive argument: motivation to comply with one’s employer, clergyman. or piano
teacher is likely to be restricted to particular sets of behavior. By the same token. “sweet-
ness” may be perceived as a good thing in certain desserts but may be evaluated negatively
when it comes to wine. In a number of examples in their recent book (e.g.. Ajzen 6i
Fishbein, 1980. p. 269). the authors do in fact ask people to evaluate behaviorally specifc
outcomes (e.g.. “making new nuclear power plants safer than present ones”) rather than
general ones (e.g.. “safety”). Furthermore. they suggest in their paper that perhaps they
should shift their emphasis “from a general to a behavioral domain level of analysis” with
respect to motivation to comply. Whether it will be possible to develop a useful measure of
MC that is not so specific as to tap behavioral intentions is still an open question.
AN EXAMINATION OF THE FISHBEIN-AJZEN MODEL 337
Bettman. J. R.. Capon. N.. & Lutz. R. J. Multiattribute measurement mudels and multiattri-
bute attitude theory: A test of construct validity. Journal ofConsumer Research. 1975.
1, I-15. b
Birnbaum. M. H.. & Mellers. B. A. Stimulus recognition may mediate exposure effects.
Journal of Personality and Social Psychology. 1979, 31, 391-394.
Calder. B. J.. & Burnkrant. R. E. Interpersonal influence on consumer behavior: An
attribution theory approach. Journal of Consumer Research, 1977. 4, 29-38.
Cohen, J.. & Cohen. P. Applied multiple reRrrssion/c,orre[atiun una~~sisjiv the hrhuGoral
s&~nc~~s. Hillsdale. N. J.: Erlbaum. 1975.
Cohen. J. B.. & Ahtola. 0. T. An expectancy X value analysis of the relationship between
consumer attitudes and behavior. In Proceedings. Second annual conference. Chicago:
Association for Consumer Research. 1971. Pp. 334-364.
Deutsch. M., & Gerard. H. B. A study of normative and informational social Influences
upon individual judgment. Joumul c!f’ Ahnormul and Social Ps~cholo,q. 19.55. 52,
629-636.
Dickson, P. R.. & Miniard, P. W. A further examination of two laboratory tests of the
extended Fishbein attitude model. Journal o/‘Cons~mer Research. 1978. 4, 261-266.
Draper, N. R.. & Smith, H. Applied regwssion una!\sis. New York: Wiley, 1966.
Dulany. D. E. Hypotheses and habits in verbal “operant conditioning.” Journal qf’Ahnor-
ma/ and Social Psychology. 1961. 63, 251-263.
Dulany, D. E. Awareness. rules. and propositional control: A confrontation with S-R
behavior theory. In D. Horton & T. Dixon (Eds.), Verbal behavior rrnd S-R behul~ior
theor!. Englewood Cliffs, N.J.: Prentice-Hall, 1968.
Fishbein. M. Extending the extended model: Some comments. In B. B. Anderson (Ed.).
Adrwzce.\ in consumer rrseclrch (Vol. 3). Chicago: Association for Consumer Research.
1976. Pp. 491-497.
Fishbein. M.. &I Ajzen. I. Eeliyf; urtitude. intention. und hehu\~ior. Reading. Mass.:
Addison-Wesley. 1975.
Fishbein. M.. & Ajzen. I. Misconceptions about the Fishbein model: Reflections on a study
by Songer-Necks. Journal oj’ Experimental Social Psyho/og~. 1976. 12, 579-584. a
Fishbein. M.. & Ajzen. I. Misconceptions revisited: A final comment. Journal of Experi-
mental Social Ps.vchology, 1976, 12, 591-593. b
French, J. R. P.. & Raven, B. The bases of social power. In D. Cartwright (Ed.). Studies in
sorid porr’rr. Ann Arbor: Univ. of Michigan Press, 1959.
Green. P. E. Ana/y;ing multivariate daru. Hinsdale, Ill: The Dryden Press. 1978.
Hays. W. L. Statistics fiv ps~cho/ogist,s. New York: McGraw-Hill. 1967.
Jaccard. J.. & Davidson. A. R. A comparison of two models of social behavior: Results of a
survey sample. Sociometry. 1975. 38, 497-517.
Johnston, J. E‘c~m<lrnetric mrfhocis. New York: McGraw-Hill, 1963.
Lutz. R. J. Changing brand attitudes through modification of cognitive structure. Journal c[f
Consumer Rr.seurch. 1975. 1, 49-59.
Lutz, R. J. Conceptual and operational issues in the extended Fishbein model. In B. B.
Anderson (Ed.), Advunce.s in consumer re,search (Vol. 3). Chicago: Association for
Consumer Research. 1976. Pp. 469-476.
Lynch. J. G.. Jr. Why additive models fail as descriptions of choice behavior. Journal of’
Experimental Sociul Pswho1og.v. 1979, 1.5, 397-417.
Miniard. P. W., & Cohen. J. B. Isolating attitudinal and normative influences in behavioral
intentions models. Journul of Murkrting Reseurch. 1979, 16. IO?- 110.
Overall. J. E.. Rc Speigel, D. K. Concerning least squares analysis of experimental data.
P.cyhoiogicul Bulletin, 1969, 72, 3 I l-322.
Roscoe. J. T. Fundamental rescurclr srcztistics for rhv hehar~iuralscirncrs. New York: Holt.
Rinehart Hc Winston. 1975.
AN EXAMINATION OF THE FISHBEIN-AJZEN MODE12 339
REFERENCE NOTES
I. Fishbein. M. Sexual behavior und propositional control. Paper presented at the Psy-
chonomic Society meetings. 1966.
2. King. G, W.. & Jaccard, J. J. The relationship between behavioral intention. attitudinul
and normative variables. Paper presented at the Speech Communication Association.
New York. 1973.
3. Glassman. M.. & Birchmore. M. The relationship het,l,een subjectiw norms and normn-
tive beliefs. Unpublished manuscript, University of Illinois. 1974.