ML20235D849
| ML20235D849 | |
| Person / Time | |
|---|---|
| Site: | Shoreham File:Long Island Lighting Company icon.png |
| Issue date: | 07/08/1987 |
| From: | Ajzen I, Fishbein M ILLINOIS, UNIV. OF, URBANA, IL, MASSACHUSETTS, UNIV. OF, BOSTON, MA |
| To: | |
| References | |
| OL-3-A-008, OL-3-A-8, NUDOCS 8709250379 | |
| Download: ML20235D849 (17) | |
Text
._
i
- c gC ICEK AJZEN t
University of Massachusetts g,{f D MARTIN FISHBEIN univers:tv of iiiinois y,3 g_ g 3
.g g v+,
n -r 4
Understanding Attitudes i
and Predicting Social Behavior i
r l
1 Prentice-Hall, Inc., Englewooc) tilffs, New Jersey 07632 i
9 v
a
[
8709250379 870708 PDR ADOCK 05000322 I
O PDR L
a i
Library of Congress Caratogsng in Pubhcarton Data Ape n, le e k.
tJnderstanding attitudes and predicting social behavior Bibliography : p. 24 8 includes index.
- 1. Attitude (Psychology)
- 2. Prediction (Psychology)
- 3. Social psychology-Addresses, essays, lectures.
1 Fishbein. Martin, joint author.
!!. Title.
BF323 C5 A37 301.1 79 26063 ISBN 0-13-936443-9 ISBN 013-936435-6 pbk.
TO OUR MOST IMPORTANT OTHERS:
RACHEL, RON, AND ELIE l
DEBORAH AND SYDNE Y l
O 1980 by Prentice Hall, Inc., Englewood Cliffs, N.J. 07632 h'/8"'
- All nghts reserved. No part of this book may be reproduced in any form or A THEORY OF by any means without permission in writing from the publisher.
C.
Intr A THEORY OF REASONED ACT Pnnted in the United States of America PLAN OF THE BOOK 10 9 8 7 6 5 4 3 2 C
Prentice Hall international. Inc., London Attitudes Prentice Ilall of-Australia Pty. Limited, Sydney Prentice Hall of Canada, Ltd., Toronto A filSTORICAL PERSPECTIVE Prentice. Hall of India Private Limited,New Delhi Prentice-Hall of Japan. Inc., Tokyo Early A ttitude Measurement,14 Early Attitude Research,16 Prentice Hall of Southeast Asia Pte. Ltd., Singapore Relation Between Attitude andi Whitehall Books Limited, Wellington, New Zealand
Attitudes and Behavior 27 retation of the other variables approach evaluation of any psychological object and we draw a clear distinction between
, behavior is moderated by other vari-beliefs, attitudes, intentions, and behaviors,in the following chapters we discu ates (Fazio & Zanna,1978; Regan &
each of these concepts and describe the relations among them,as postu
.vior relations of moderate rnagnitude ur theory of reasoned action. In contrast to the currently accepted view that ience with the attitude object or with there is no close link between attitude and behavior, our approach suggests tha
- experience, the relation between atti-appropriate measures of attitude are strongly related to action. Like Thurstone,
'ely weak. Other moderating variables Doob, and Campbell we take the position that attitudes toward an object ca considered presence of others, posses, predict only the overall pattern of behavior; they are oflittit value if we are in.
avior, internal consistency of the atti.
terested in predicting and understanding some particular action with respect to is held, and occurance of unforeseen the object. To predict a single behavior we have to assess se person's attitude ese fcctors have been found to moder.
toward the behavior and not his attitude toward the target at which the behavior ematic fashion, they have often been is directed. In other words, according to our approach any behavioral criterion udes are found to predict behavior in can be predicted from attitude-be it a single action or a pattern of behavior-nificance of moderating variables and Provided that the measure of attitude corresponds to the measure of bebavior.
m will be considered in our discussion A recent review of the attitude behaviorliterature (Ajzen & Fishbein,1977)sup
),
ports this position.
I The attitude behavior problem has led not only to a re examination of the attitude concept but also to a growing recognition that,in marked contrast to SION the amount of effort invested in the development of reliable and valid attitude measures, very little attention has been paid to the assessment of behavior (see Fishbein,1967a,1973; Fishbein & Ajzen,1974; Schuman & Johnson,1976) i a giowing interest in the relationship Clearly, adequate conceptualization and measurement of behavior is essential for n,1976; Calder & Ross,1973; Liska, an understanding of the attitude behavior relation. Accordingly, we begin our mson,1976). The attitude behavior discussion of the theory of reasoned action in chapter 3 by reviewing the natur iractical implications in that it has led and measurement of behavioral criteria.
measurement of attitudes. Although it s, a trend to return to the unidimen-or affect with respect to a psycho-
,1976; Schuman & Johnson,1976),
addition to affect, attitudes also con-n fact, most investigations concerned io distinctions among beliefs, feehngs, s-and sometimes even overt actions-t's " attitude," and measures of these i y general agreement that attitude,no
.my factors that influence behavior. it
,ccurately we have to take additional dent contributors to behavior or as nship. While re. affirming the impor-modates the findings oflow or incon.
and behavior.
1 that are clearly at variance with this
- t the term " attitude" to a person's
50 A Theory of Reasoned Action ous events. Since most investigators make no distinction between attitudes and To summarize briefly, we haw intentions (see chapter 2), these variables would also be expected to moderate tion of behavior, intentions that j the intention-behavior relation, prior to observation of the behat in marked contrast, we are claiming that intentions should always predict be-intention should be taken as cloe havior, provided that the measure of intention corresponds to the behavioral is sometimes possible to improve j criterion and that the intention has not changed prior to performance of the These conditional intentions tah behavior. From our point of view,other variables may appear to have moderating the investigator, that might proe effects on the strength of the intention behavior relation if the second of these long range prediction from intes requirements is not met. This can easily be seen in the case of unforeseen extra-level, even when the measure ofic neous events. Clearly, when such events occur, they will tend to change inten-individual behavior.
tions and thus re:it.ce the relationship between a previous measure ofintention and subsequent behavior. It follows that the intention behavior relation will tend to be strong in the absence of unexpected events but weak in their presence.
PREDICS However, as noted above,if the intention is measured after the extraneous events have occurred (or if those events are taken into account) a strong intention.
behavior relation will be obtained.
We noted in chapter 3 that inves Similar considerations apply to the other " moderating variables." Take, for outcomes rather than behaviors, example, the case of direct experience with the behavioral target. An intention cussed in chapter 9. Unlike bA formed without such direct experience may change greatly when the person is person's volitional control. For es actually confronted with the target of the behavior. In contrast, a person who the ne'xt two years depends not oc has interacted with the target in the past will have developed more realistic ex-as her husband's behavior, the avd pectations, and his intention is likely to remain relatively stable.'
couple's fecundity.
A concrete example may help illustrate this point. Imagine.i young woman It may appear that the simpt who is quite attractive but who, on first encounter, appears a bit reserved and measure the woman'sintention te rejecting. Suppose Mr. A has mteracted with her on several past occasions and ever, that use of such a measure (
has come to realize that her coldness is only a front. Mr. B on the other hand is in question will result from the pe only superficially acquainted with her. Let us assume that because both men are tention to produce the outcome e attracted to the woman, each forms the intention to ask her out. In their con.
When these assumptions are met, versations with her prior to extending an invitation each man is confronted with predict its occurrence. When they her typical reserve. Their reactions,however,may differ greatly.Mr. A expects a
- tionship, certain coolness, is unperturbed, and carries out his intention. Mr. B, who does A more promising approach is a not realize that the reserve is only superficial,is taken aback,perhaps offended, control the outcome. To predict t and quickly changes his intention. Mr. B's behavior is consistent with his new in-son's intentions to perform each c tention not to ask the woman for a date, but it is unrelated to his initialinten-ence or absence of the other factog tion as measured on a previous occasion.
It can be seen that direct experience may help preserve a stable intention and hence a strong intention behavior relation. Note,however, that direct experience SUMMARS need not necessarily lead to the development of more realistic expectations.
Direct experience will therefore not always influence the stability ofintentions (or attitudes) and hence will not always moderate the strength of the intention-In this chapter we have considerG behavior relationship.
tions to predict behavior. We ham
'The same arguments apply to the stability of a person's attitude toward the target or to be the immediate antecedents a toward a behav or with respect to the target. Here, too, direct experience may influence tion and behavior depends on ta the stability of the attitude and hence moderate its relation with behavior, to correspond to the behavioral e
I J
Predicting Behavior frcm Intention $1 distinction between attitudes and To summarize briefly, we have seen that in order to ensure accurate predic.
dd also be expected to moderate l
tion of behavior, intentions that are not stable have to be measured immediate entions should always predict be-prior to observation of the behavior. When this cannot be done, the measure of n conesponds to the behavioral intention should be taken as close in time as possible to the behavior. Further,it ged prior to performance of the is sometimes possible to improve prediction by measuring conditionalintentions.
These conditional intentions take into account extraneous events, foreseen by
- s may appear tohave moderating I
>r relation if the second of these the investigator, that might produce changes in intentions. Finally, we saw that long range prediction from intentions will usually be accurate at the aggregate.
1in the case of unforeseen extra-
. they will tend to change inten-level, even when the measure ofintention does not permit accurate prediction of individual behavior.
a previous measure ofintention I
ntion behavior relation will tend mts but weak in their presence.
iured after the extraneous events j
PREDICTION OF OUTCOMES to account) a strong intention.
}
noderating variables." Take, for We roted in chapter 3 that investigators are sometimes mterested in predicting behavioral target. An intention outcomes rather than behaviors. The case of predicting weight loss will be dis.
mge greatly when the person is cussed in chapter 9. Unlike behaviors, outcomes are not completely under a vior. In contrast, a person who person's volitional control. For example, whether a woman will have a child in iva developed more realistic ex-the next two years depends not only on her own actions but also on such factors
'latively stable.'
as her husband's behavior, the availability of contraceptives er abortions, and the point, imagine a young woman couple's fecundity, it may appear that the simplest way to predict this outcome would be to iter, appears a bit reserved and i r on several past occasions and measure the woman's intention to have a child in the next two years. Note how-
'ont. Mr. B on the other hand is ever, that use of such a measure is based on two assumptions: that the outcome ume that because both men are in question will result from the performance of certain behaviors and that the in-n to esk her out. In their con-tention to produce the outcome willlead to the performance of those behaviors, m each man is confronted with When these assumptions are met, then the intention to produce an outcome will differ greatly. Mr, A expects a predict its occurrence. When they are not met, we cannot expect a strong rela-tionship.
- his lateation. Mr. B, who does l rken aback, perhaps offended, A more promising approach is to identify the behaviors and other factors that er is consistent with his new in-control the outcome, To predict its occurrence one could then measure the per-s unrelated to his initialinten-son's intentions to perform each of the identified behaviors and assess the pr ence or absence of the other factors, preserve a stable intention and towever, that direct experience if more realistic expectations.
SUMMARY
AND CONCLUSION nce the stability ofintentions
. 2 the strength of the intention-In this chapter we have considered some of the issues involved in using inten-tions to predict behavior. We have noted that although intentions are assumed 2n's arrirude toward the target or I
5*
- I"
, direct experience rnay innuence tion and behavior depends on two factors: First, the measure ofintention has s ofth behavior.
to correspond to the behavioral criterion in netion, target, context, and time; 2
e maie a
g tion of behavior provided by the the tent area to another, the same basic involved. Moreover, we tried to she i
the theory, such as demographic cD 1
tional measures of attitudes toward @
effects on beliefs, attitude toward &
The theory of reasoned action thus can help explain many of the inconsis At this point, some words of caw there are likely to be some human theory of reasoned action. Among th performance of well-learned skills, sua in this book we have described an approach to the prediction and understanding a car. There may also be certain indivi of human behavior and we have illustrated its application in a variety of behav-some people may arrive at their dect ioral domains. Basic to this approach is the view that people use the information evidence suggests, however, that the available to them in a reasonable manner to arrive at their decisions.Thisis not with respect to most sociai behaviors.
to say that their behavior will always be reasonable or appropriate from an ob.
pear unplanned turn out, on closer en jective point of view. People's information is often incomplete and at times also in chapter 4 that low intention behas incorrect. But we would argue that a person's behavior follows quite logically intentions rather than evidence for ung and systematically from whatever information he happens to have available.
A related issue concerns use of the Within our approach, this information constitutes the person's beliefs. Of the tion processing. Although we take thc many different kinds of beliefs a person holds, we have identified two types as and subjective norms'and that them underlying any given action: behavioral beliefs and normative beliefs. We have mean to imply that prior to perform!
tried to show that systematic processes link these beliefs to behavior by way of atically scrutinize the determinants og attitude toward the behavior, subjective norm, and intention. Each stage in the cesses involved as largely automatic e sequence follows reasonably from the preceding stage, For example, it is rea-become fully aware of these processsez sonable to feel positively about performing a behavior if you believe that its per-toward smoking cigarettes. As the pem formance will lead to more good than bad outcomes. It is also reasonable to feel advantages or disadvantages, she simu3 social pressure to not perform a behavior if you believe that people with whom a positive or negative attitude towaa you are motivated to comply think you should not perform it. Finally,it is rea-opportunity to smoke, this favorable sonable to weigh your personal feelings (attitude) and the perceived social pres-directly aroused without any need os sure (subjective norm) in arriving at and carrying out your intention. Taken review her beliefs about the behaviorat together, the processes involved in this sequence comprise a theory ofreasoned attitude influences her decision to sms action.
explain her intention or behavior is t in part 2 of this book we showed that this theory can account for behavior in feelings toward smoking,of the socialg I
such diverse areas as weight reduction, occupational choice, family planning, her behavioral and normative beliefs et brand choice, voting in American and British elections as well as in referenda, sion.2 and voluntary treatment programs. The studies described,however, served merely It may also be usefulin this concle as illustrations. The theory has been applied in other domains, including use of public transportation (Thomas,1976; Thomas, Bull, & Clark,1978), reenlist-For detailed dis:ussions of attitude foGD 3
ment in military organizations (Hom,Katerberg,& Hulin,1978;Keenan,1976),
@en (1975, chap. 6).
church attendance, (King,1975), drug abuse (Pomazal & Brown,1977), and j C ["'rtEn"una$are$f 8
blood donation (Pomazal & Jaccard,1976). Althour. the substantive e,olana*
factors) that influence their beliefs, sttitude3s 244
~ _ _ _
Some Final Comments 24$
tion of behavior provided by the theory of reasoned action varies from one con-tent area to another, the same basic constructs and psychological processes are involved. Moreover, we tried to show that the effects of variables external to the theory, such as demographic characteristics, personality traits, and tradi-tional measures of attitudes toward objects, can be explained by examining their effects on beliefs, attitude toward the behavior, subjective norm,and intention.
The theory of reasoned action thus provides a comprehensive framework that can help explain many of the inconsistent findings in applied social research.
At this point, some words of caution are in order. It should be clear that there are likely to be some human behaviors that cannot be explained by. a i
theory of reasoned action. Among these behaviors are emotional outbursts and performance of well-leamed skills, such as turning the pages of a book or driving pproach to the prediction and understanding a car. There may also be certain individuals for whom the theory does not apply; istrated its application in a variety of behav-some people may arrive at their decisions in different ways. The accumulating h is the view that people use the infonnation evidence suggests, however, that the theory is useful for most individuals and anner to arrive at their decisions.This is not with respect to most social behaviors. in our opinion,many behaviors which ap-
/s be reasonable or appropriate from an ob-pear unplanned turn out, on closer examination,to be quite intentional.We saw mation is often incomplete and at times also in chapter 4 that Icw intention behavior relations may often reflect changes in t a person's behavior follows quite logically intentions rather than evidence for unplanned behavior.
formation he happens to have available.
A related issue concerns use of the theory as a description of human informa-
.ation constitutes the person's beliefs. Of the tion processing. Although we take the position that beliefs determine attitudes rson holds, we have identified two types as and subjective norms and that these in turn influence intentions, we do not ioral beliefs and normative beliefs. We have mean to imply that prior to performing each and every action, people system-sses link these beliefs to behavior by way of atically scrutinize the determinants of their behavior. Rather, we view the pro-
- tive norm, and intention. Each stage in the cesses involved as largely automatic or implicit, and only in rare cases do we the preceding stage. For example, it is rea-become fully aware of these processses. Consider,for example,a person's attitude
'orming a behavior if you believe that its per-toward smoking cigarettes. As the person forms beliefs that smoking has certain m bad outcomes. it is also reasonable to feel advantages or disadvantages, she simultaneously and automatically also acquires tavior if you believe that people with whom a' positive or negative attitude toward smoking.3 When confronted with the you should not perform it. Finally,it is rea-opportunity to smoke, this favorable or unfavorable feeling toward smoking is ings (attitude) and the perceived social pres-directly aroused without any need on the part of the person to systematically at and carrying out your intention. Taken review her beliefs about the behavior and, together with her subjective norm, the his sequence comp 6se a theory ofreasoned attitude influences her decision to smoke or to not smoke. Only when asked to explain her intention or behavior is she lin ely to become fully aware of her that this theory can account for behavior in feelings toward smoking,of the social pressure to smoke or to not smoke, and of
- tion, occupational choice, family planning, her behavioral and normative beliefs underlying these determinants of her deci-and British elections as well as in referenda, sion.2 The studies described however, served merely It may also be useful in this concluding chapter to address a different issue n applied in other domains, including use of 76; Thomas, Bull, & Clark,1978), reenlist-3ror detailed dacussions of attitude forrnation, see Fishbein (1967b) and Fishbein and 1, Katerberg, & Hulin,1978; Keenan,1976),
Ajren (1975, chap. 6).
2 IruE abuse (Pomazal & Brown,1977)' and in c ntrast, recent research (Nisbett & Wilson,1977) suggests that even when asked, people are often unaware of external variables (dernographic characteristics, situational 1,1976). Although the substantive explana-factors) that influence their behefs, attitudes, intentions, and actions, 244
'"d._s 1
246 Applications j
that is frequently raised with respect to the theory of reasoned action.The issue motivation to comply with a ghen rc is typically viewed in terms of general versus specific attitude measures, it is
.j under consideration. We have previG l
sometimes mistakenly assumed that the theory of reasoned action involves highly specific behavioral prescription (F12 l
specific attitudes that are relevant only for a single, unique behavior. Since the such a measure of motivation to em theory does refer to attitude toward a behavior (rather than attitude toward an j
of the normative component which object or target), this impression is understandable. We have tried to make it however, that motivation to compR clear, however, that the theory of reasoned action can be applied at any level that is, at the level of complyir:g wid of generality. The behavioral criterion selected by the investigator determines example,instead of asking people b how specific or general his measure of attitude (as well as his measures of beliefs, say, their friends think they should i
intention, and subjective norm) has to be. If, for whatever reason,he is interested
[
want to do what their friends this in predicting and understanding a single action, directed at a certain target,in a planning or to politics. Here, too,1 specific context, and at a given time, then of course his measure of attitude must other questions concerning the concG correspond to his criterion in all of these elements. Usually, however, our in-Some theorists (e.g., Schwartz e terest is in much more global behaviors which are general at least with respect to have argued that in addition to pec context and time and often with respect to target or action as well. Thus, we account the individual's own feella may want to understand the factors determining cigarette smoking (irrespective perform the behavior under investig of the brand of cigarettes, the context, the time)or discrimination against mem-of moral norm be added as a third ce bers of a minority group regardless of the specific form this discrimination takes.
Similarly, Triandis (1967,1977) hc Here very specific attitudes would be inappropriate instead, we would have to havior that may also be classified o assess attitudes toward smoking cigarettes or toward the minority group in ques-spects, the theory is quite similar to e l
tion. Similarly, intentions, subjective norms, behavioral and normative beliefs noted himself,it makes no attempt "
I would also have to correspond in their levels of generality to the behavioral cri-toward a behavior,it includes a sepe l
terion as defined by the investigator. The theory of reasoned action thus requires of performing the behavior. Mores not that we assess specific attitudes but rather that our measures of attitude social determinants, but within his always correspond to our measures of behavior, measurement of norms, roles, self-es Finally, we would like to make it clear that the theory formulated in this agreements.
book is still in the process of development. Many years ofintensive work on atti.
We realize, of course, that future tude theo y and measurement have given us a good understanding of the ways sider additional factors in our atteam attitudes are formed and of how they can best be measured. Many questions, this book we have tried to show hs however, remai:i with respect to the normative component in our theory, attitudes and subjective norms are sy First, we are not at all certain that the measure of subjective norm proposed at the present time, we see no need te in this book is the best way of directly assessing perceived social pressure to per-form or to not perform a behavior. Although we are convinced that perceived l
l social pressure must be taken into account in order to explain social behavior, al-ternative ways of assessing that pressure might be developed. For example, re-search currently in progress is exploring the utility of asking people whether their important others perform the behavior under consideration instead of (or in addition to) asking them whether these important others think they should per-form the behavior. The same question also apphes at the level of normative be-hefs, that is, the perceived normative prescription of a given individual or group.
3 Closely related to the latter question, there are many unresolved issues with
[*,ChN*s,e stajem' f h ng na
,, ti respect to the concept of motivatior; to comply. For examp'e,in this book we the third component from the theory sins have conceptualized and measured motivation to comply at a very generallevel:
way of measunng behavioralintention.
Some Final Comments 247 i the theory of reasoned action. The issue motivation to comply with a given referent irrespective of the particular behavior l versus specific attitude measures. It is under consideration. We have previously argued against tying this concept to a theory of reasoned action involves highly specific behavioral prescription (Fishbein & Ajzen,1975, p. 306) since we see for a single, unique behavior. Since the such a measure of motivation to comply as playing a role similar to the weight behavior (reher than attitude toward an of the normative component which is also behavior-specific. It could be argued, iderstanda We have tried to make it however, that motivation to comply is best measured at an intermediate level, soned action can be applied at any level that is, at the level of complying with a referent within a behavioral domain. For selected by the investigator determines example,instead of asking people how much,in general, they want to do what, tritude (es well as his measures of beliefs, say, their friends think they.hould do, they could be asked how much they
- e. If, for whatever reason,he is interested want to do what their friends think they should do when it comes to family r action, directed at a certain target,in a planning or to politics. Here, too, research is in progress to answer these and
, rn of course his measure of attitude must other questions concerning the concept of motivation to comply.
' tese elements. Usually, however, our in-Some theorists (e.g., Schwartz & Tessler,1972; Pomazal & Jaccard,1976) l which are general at least with respect to have argued that in addition to perceived social pressures, we must take into ect to target or action as well. Thus, we account the individual's own feeling of moral obligation or responsibility to termining cigsrette smoking (irrespective perform the behavior under investigation. They have suggested that a measure the time) or discrimination against mem-of moral norm be added as a third component in the theory of reasoned action.8 ie specific form this discrimination takes.
Similarly, Triandis (1967,1977) has developed a theory of interpersonal be.
1 appropriate. Instead, we would have to havior that may also be classified as a theory of reasoned action. In some re-
- es or toward the minority group in ques.
spects, the theory is quite similar to our own, but as Triandis (1977,pp.231232) iorms, behavioral and normative beliefs noted himself,it makes no attempt to be parsimonious. In addition to attitude levels of generality to the behavioral cri-toward a behavior,it includes a separate measure of the perceived consequences e theory of reasoned action thus requires of performing the behavior. Moreover, Triandis recognizes the importance of at rather that our measures of attitude social determinants, but within his theory this construct involves the separate rhavior.
measurement of norms, roles, self concept, moral norm, ideals, and contractual
- lear that the theory formulated in this agreements.
nt. Many years ofintensive work on atti.
We realize, of course, that future research may demonstrate the need to con-en us a good understanding of the ways sider additional factors in our attempts to predict and understand behavior,in j can best be measured. Many questions, this book we have tried to show,however, that, when appropriately measured, mative component in our theory.
attitudes and subjective norms are sufficient to predict intentions, and, at least he measure of subjective norm proposed at the present time, we see no need to expand our theory of reasoned action.
assessing perceived social pressure to per-though we are convinced that perceived
.nt in order to explain social behavior, al-ce might be developed. For example, re-the utility of asking people whether their j
ir under consideration instead of (or in 2 important others think they should per-also applies at the level of normative be-description of a given individual or group.
1, there are many unresolved issues with in fact, the origmal statement of our theory (Fishbein,1967a) included personal nor-mative behefs in addition to attitude and social normative behefs. Research led us to drop
> comply. For example,in this book we the third component from the theory since it seemed to be little more than an alternative ivation to comply at a very generallevel:
way of measuring beWral intention.
I I
l L.
-____.______.______.____m_________________
46 A Theory of Reasoned Action (i.e., subjective probabilities) with respect to each of the action alternatives. This A measure of intention that corra set of behavioral intentions can also be used to predict the choice. Imagine,for as follows. M :neasue the responda example, that a person had indicated the following subjective probabilities prior single actions corpprismg the behavis to the 1976 presidential election.
each intention correspond to those c going to cbserve whether a person ha I will vote for Gerald Ford.
30% chance Period, we would assess his intention t I will vote for Jimmy Carter.
40% chance Period. We can then compute an ins I will vote for some other candidate.
3e chance described in chapter 3 for computing 1 I will not vote.
25% chance in sum, we aw seen Gat H h q pondence between intention and beu It stands to reason that the person will perform that alternative to which he de or a MaM categg M M assigned the highest subjective probability; in this case,he is most likely to have
'**U "'**'8'*'C " " '"
voted fot I!mmy Carter.
l Note that a person's intention to perform one of the behavioral alternatives l
cannot always be used to predict his choice.2 For example, a person who indi-STABILITY (
cates a 40% chance that he will buy Shell gasoline the next time he needs gas I
)
may in fact do so if the probabilities for any other brand are even lower, but he may buy Exxon gasoline if his subjective probability for Exxon is, say,45%.
We have argued that a measure ofinc An analogous problem arises in the prediction of behavioral magnitudes or will lead to accurate prediction. TM frequencies. Knowing that a person has a 40% likelihood of donating $100 to any attempt to understand and infly the United Fund tells us little about the amount of meney he will actually donate relation between mtention and behat because it does not tell us which amount of money has the highest subjective tenti ns is both necessary and suf2 probability. To carry this one step further, measuring a person's behavioral in.
human action. However, a measure o tention "to donate money to the United Fund" is also unlikely to predict the dictor of behavior. It should be obvioy i
I amount of money actually donated. As a general rule, then, a person's intention a measure oNntenh tab some a, to perform a given behavior cannot be used to predict the extent, magnitude,or differ from the mtention at the time frequency of that action, the time interval is, the greater the lih produce changes in intentions. Generd Correspondence for Behavioral Categories interval, the less accurate the predicti lower the observed relation is betweea Our discussion in chapter 3 showed that measures of behavioral categories repre-sent the degree to which an individual engages in a general class of behaviors. We Predicting Behavior also noted that asking a person to classify himself as performing or not perform-ing the behavi0ral category (e.g., dieting) does not provide a meaningful classifi-Since intentions can change over tims catiott, since sud judgments are arbitrary. lt follows that a measure of intention as close as possible to the behavioral 4 to diet or to pe' form some other general category may not be a good predictor Prediction. In many instances, howem of the behavio.alindex computed on the basis of a set of single actions.3 tical value to measure the intention la for example, that we are trying to prea a
One important exception is the smgle action critenon which may be viewed as a choice field. We could hardly approach soldia between performmF and not performmg the behavior. Here, a person's intention to perform their mtentions. In other cases,it ma, the behavior is the complement of his intention to not perform the behavior, That is, the a
probabihty of performing the behavior is one minus the probabihty of not performing the make predictions months or years in a behaviot, and knowmg one probabihty is sufficient to predict the choice.
not concerned with the behavior of f if the behavioral criterion is a report of whether the person views himself as dictmg or
.ecting of forecasting behavioral trende not dictmg, this measure of intention should provide accurate prediction.
tion. For example, market researchers ;
l
\\
Predicting Behavior from Intention 47 i o each of the action alternatives. This d to predict the choice. Imagine, for A measure of intention that corresponds to a behavioralcategory is obta illowing subjective probabilities prior as follows. We measure the respondent's intentions with respect to each i
single actions comprising the behavior index, making sure that the eleme each intention correspond to those of the behaviors. For example,if we we
, 0% chance going to observe whether a person had snacked between meals over a two
- _0% chance period,we would assess his intention to snack between meals in that two-m period. We can then compute an intentionalindex in a manner similar to tha S.fechance 5% chance described in chapter 3 for computing the behavioralindex.
In sum, we have seen that it is important to ensure a high degree of corre
~
pondence between intention e.nd behavior, whether the criterion is a si
' rform that alternative to which he tion or a behavioral category. Lack of correspondence on any of the four e n.this case, he is most likely to have (action, target, context, and time) can reduce the accuracy of prediction, j (one of the behavioral alternatives For example, a person who mdi.
STt,BILITY OF INTENTIONS asolina the next time he needs gas other brand are even lower, but he ability for Exxon is, say,45%.
ction of behavioral magnitudes or We have argued that a measure of intention which corresponds to the b j % likelihood of donating $100 to will lead to accurate prediction. This argument has important implications any attempt to understand and influence human behavior. Given a one to one tt of money he will actually donate relation between intention and behavior, knowledge of the determinants o money has the highest subjective tentions is both necessary and sufficient for understanding and influe j ieasuring e person's behavioral in-human action. However, a measure of intention will not always be a id"is also unlikely to predict the ral rule, then, a person's intention dictor of behavior. it should be obvious that intentions can change over a measure ofintention taken some time prior to observation of the beh predict the extent, magnitude, or differ from the intention at the time that the behavior is observed.
the time interval is, the greater the likelihood is that events will occur i
{
produce changes in intentions. Generally speaking, therefore,the longer interval, the less accurate the prediction of behavior from intention, t lower the observed relation is between intention and behavior.
es of behavioralcategories repre-n a general class of behaviors. We Predicting Behavior ilf as performing or not perform-tot provide a meaningful classifi-lows that a measure ofintention Since intentions can change over time,it is important to measure the ary may not be a good predictor as close as possible to the behavioral observation in order to obtain an f a set of single actions.:
Prediction. In many instances, however, it may be neither feasible n ticai value to measure the intention in close proximity to the behavio for example, that we are trying to predict the behavior of a soldier o ion which may be viewed as a choice lere e person's intention to perform i eld. We could hardly approach soldiers in the midst of a battle and I perform the behavior, That is, the their intentions. In other cases,it may be important, for planning pu UQ not performing the make predictions months or years in advance. Long range predic 3
not concerned with the behavWr of any given individual but rather with j
e person views himself as dieting or jecting or forecasting behavioral trends in relatively large segmentl 2 rate prediction, tion. For example, market researchers are interested not so much in the 3
)
1
i t
48 A Theory of Reasoned Action that a given individual will buy a car in a given calendar year but in the number new sample of individuals) in order te or proportion of people who will buy cars in that year. By the same token, the have taken place. For purposes of 113 military is concerned not with predicting the enlistrnent behavior of a given indi.
If the manufacturer ha'd foreseen the g vidual but rather with the number of volunteers that will enlist at some future gasoline, he could have obtained an i
- time, conditional upon the occurrence of thf The distinction between predicting behavior at the level of the individual and could also have asked his respondents at the aggregate level is important because aggregate intentions sre apt to be car if the price of gasoline increased D much more stable over time than are individual intentions. A great variety of crease in the price of gasoline, this e eveats can produce changes in the intentions of individuals. Among such unex.
accurate prediction of behavior than tDi pacted events are sudden illness; a fortuitous inheritance; injury; natural disasters, There are mar.y long range predict such as tornadoes, floods, or earthquakes; loss of job; unexpected pregnancy; conditional intention measures. To rs and economic recession. Similarly, if a person's intention is based on the expec.
of air travel could be improved by &
tation that another person will behave in a certain way or on the expectation creased air fares, and prediction of tr9 that some event will occur, and the expectation is not confirmed, this informa.
by taking into account events that ra tion may also lead to a change in intention, change in government, re valuation of 6 Since such idiosyncratic events are likely to balance out, intentions at the Even relatively short. range predicti eggregate level will tend to be relatively stable. For example, one person may tional intentions. To give just one en have intended to buy a new car but, because an expected rise in salary does not might be predicted with greater acce occur, she changes her intention. Another person might be involved in an acci.
upon the weather.
dent and, although he did not intend to buy a new car, he finds himself in the Note that in all of these examplea position where he has to buy one. Although the behaviors of these individuals general public and thus also to the fg could not have been predicted, the changes in intention would have had no effect sideration of such events as illness,los on the accuracy of the aggregate prediction. In fact, there is considerable evi.
also improve prediction, they are of13 dence that ever. when individual predictions are relatively poor, prediction of be.
person's behavior s To.uid out if som havior from irtention at the aggregate level is often remarkably accurate (see unexpected compary one would have chapters 11 and 12).
is possible to re. interview respondents :
One exception to this rule is the case of an external event that changes the in.
havioral observation,it is also possible tentions of a large proportion of the population in the same direction. For ex.
at that time. Since this more recent I ample,large increases in air fares tend to reduce people's intentions to travel by been influenced by these events, as e plane. Similarly, the outbreak of war in the Middle East greatly reduced people's occurred,it will provide a more accum intentions to visit that area.
tion.'
Conditional intentions. To make accurate long range predictions (at either the Moderating Effects of Other Variables aggregate or the individual level) we could try to identify likely extraneous events whose occurrence would change intentions. Consider, for example, an In chapter 2 we mentioned the argums automobile manufacturer who is planning his production for the coming year, of the attitude behavior relationship 12 A sample of potential car buyers is interviewed in August, and their intentions direct experience with the attitude o%
to buy the car in question during the followmg calendar year are assessed. Note of skills required to perform the behavt that the manufacturer is trying to predict behavior 4 to 16 months in advance.
While many different factors may change any given individual's intention, one "The conditional intention measure not d"
potential extraneous event might influence the intentions of a large number of
- "s" " ' ',
7,' g,,"p"l people-namely, an increase in the price of gasoline-behavior, that is, at the aggregate, as oppod Assume that a large increase in the price of gasoline occurred in January.
- Although information about the occura The manufacturer would be well advised to reinterview his respondents (or a tion of a person's behavior,it rr.ay help exp1
Predicting Behavior from Intention 49 en calendar year but in the number new sample of individuals) in order to assess any changes in intentions that ma i that year. By the same token, the enlistment behavior of a given indi.
have taken place. For purposes of illustration, imagine that this is impossible.
sers that will enlist at some future If the manufacturer had foreseen the possibility of a large increase in the pr gasoline, he could have obtained an additional measure of intention that was at at the level of the individual and conditional upon the occurrence of this event. That is,in the initialinterview,he iggregate intentions are apt to be could also have asked his respondents how likely it was that they would buy t
' jual intentions. A great variety of car if the price of gasoline increased by 20% or more. In the case of such an in-of individuals. Arnong such unex-crease in the price of gasoline, this conditional intention would lead to more theritance ; inj ury ; na tu ral disasters, accurate prediction of behavior than the original measure ofintention.'
ss of job; unexpected pregnancy; There are many long range predictions that could be improved by means of
's intention is based on the expec-conditional intention measures. To return to the previous examples, prediction ertain way or on the expectation of air travel could be improved by measuring intentions conditional upon in-m is not confirmed, this informa-creased air fares, and prediction of travel to foreign countries could be improved by taking into account events that might occur in those countries (e.g., war, a to bilance out, intentions at the change in government, re valuation of currency).
le. For example, one person may Even relatively short range predictions could be improved by means of condi-in expected rise in salary does not tional intentions. To give just one example, attendance at concerts or theaters
- on might be involved in an acci-might be predicted with greater accuracy by measuring intentions conditional upon the weather.
new car, he finds himself in the i he behaviors of these individuals Note that in all of these examples we considered only events known to the itention would havehad noeffect general public and thus also to the investigator. Although it is true that con-
'n fact, there is considerable evi-sideration of such events as illness, loss of job, and unexpected company could l relatively poor, prediction of.be-also improve prediction, they are oflittle practical value for the prediction of a person's behavior often remarkably accurate (see s To find out if somebody was taken ill, lost his job, or had unexpected company one would have to observe or interview him. Clearly,ifit l
is possible to re interview respondents to obtain this information prior to the be-l :ternal event that changes the in.
havioral observation,it is also possible to simply re assess the person's intention
) in the same direction. For ex.
I r people's intentions to travel by at that time. Since this more recent measure ofintention would have already l lie East greatly reduced people's been influenced by these events, as well as by any other events that may hav occurred,it will provide a more accurate prediction than any conditionalinten-tion.'
l range predictions (at either the l / to identify likely extraneous Moderating Effects of Other Variables l ms. Consider, for example, an in chapter 2 we mentioned the argument voiced in recent years that the str j roduction for the coming year.
) in August, and their intentions of the attitude behavior relationship is moderated by "other variables," such as I calendar year are assessed. Note direct experience with the attitude object, prese.we of other people, possessi ior 4 to 16 months in advance.
of skills required to perform the behavior, and occurrence of unforeseen extrar,e-1 iven individual's intention, one
'ne conditional intention measure not only improves prediction,it also can 9tentions of a large number of
- * "s u acturer in his initial decisions concerning production of Snat.
C' They may, however, be useful for predicting the percentage of people who perfonn the behavior, that is, at the aggregate, as opposed to the individual, level, gasoline occurred in January.
i 4
6 tierview his respondents (or a Although information about the occurrence ofidiosyncratic events does not aid predic.
l tion of a person's behavior,it may help explain that behavior after the fact.
l I
}
50 A Theory of Reasoned Action ous events. Since most investigators make no distinction between attitudes and To summarize briefly, we have sees intentions (see chapter 2), these variables would also be expected to moderate tion of behavior, intentions that are nei the intention behavior relation.
prior to observation of the behavior.W In marked contrast, we are claiming that intentions should always predict be-intention should be taken as c;ose in tk havior, provided that the measure M inwion corresponds to the behavioral
{
is sometimes possible to improve predig criterion and that the intention has not changed prior to performance of the These conditional intentions take inte behavior. From our point of view,other variables may appear tohave moderating the investigator, that might produce ch effects on the strength of the intention behavior relation if the second of these long-range prediction from intentions <
requirements is not met. This can easily be seen in the case of unforeseen extra-level, even when the measure ofintentis neous events. Clearly, when such events occur, they will tend to change inten.
individual behavior, tions and thus reduce the relationship between a previous measure ofintention and subsequent behavior. it follows that the intention behavior relation will tend to be strong in the absence of unexpected events but weak in their presence.
PREDICTION (
However, as noted above,1f the intention is measured after the extraneous events have occurred (or if those events are taken into account) a strong intention-i behavior relation will be obtained.
We noted in chapter 3 that investigator i
Similar considerations apply to the other " moderating variables." Take, for outcomes rather than behaviors. The cc example, the case of direct experience with the behavioral target. An intention cussed in chapter 9. Unlike behaviors, I
formed without such direct experience may change greatly when the person is person's volitional control. For example l
actually confronted with the target of the behavior. In contrast, a person who the next two years depends not only on; I
has interacted with the target in the past will have developed more realistic ex-as her husband's behavior, the availabilig I
pectations, and his intention is likely to remain relatively stable.'
couple's fecundity.
A concrete example may help illustrate this point. Imagine a young woman it may appear that the simplest wp l
who is quite attractive but who, on first encounter, appears a bit reserved and measure the woman's intention to have (
rejecting. Suppose Mr. A has interacted with her on several past occations and ever, that use of such a measure is basa has come to realize that her coldness is only a front.Mr. B on the other hand is in question will result from the performt only superficially acquainted with her. Let us assume that because both men are l
tention to produce the outcome willles attracted to the woman, each forms the intention to ask her out, in their con.
When these assumptions are met,then d versations with her prior to extending an invitation each man is confronted with predict its occurrence. When they are a her typical reserve. Their reactions,however,may differ greatly.Mr. A expects a
- tionship, certain coolness, is unperturbed, and carries out his intention. Mr. B, who does A more promising approach is to ides j
not realize that the reserve is only superficial,is taken aback,perhaps offended, control the outcome. To predict its ocs and quickly changes his intention.Mr. B's behavior is consistent with his new in-son's intentions to perfcan each of the tention not to ask the woman for a date, but it is unrelated to his initialinten-ence or absence of the other factors.
tion as measured on a previous occasion, it can be seen that direct experience may help preserve a stable intention and l
hence a strong intention behavior relation. Note,howeytt, that direct experience
SUMMARY
AN3 l-need not necessarily lead to the development of more realistic expectations.
Direct experience will therefore not always influence the stability ofintentions (or attitudes) and hence will not always moderate the strength of the intention-In this chapter we have considered som behavior relationship.
tions to predict behavior. We have nots
'The same arguments apply to the stability of a person's arrirude toward the tarpt or e ak anWebs d ac@
toward a behavior with respect to the tartet. Here too, direct experience may influence tion and behavior depends on two facte the stabibty of the attitude and hence moderate its relation with behavior, to " correspond to the behavioral criteri3 l
)
Predicting Behavior from Intention 51 no distinction between attitudes and would also be expected to moderate To summarize briefly, we have seen that in order to ensure accurate predic tion of behavior, intentions that ar; not stable have to be measured immedia t intentions should always predict be-prior to observation of the behavior. When this cannot be done, the measure of ntion corresponds to the behavioral intention should be taken as close in time as possible to the behavior. Further,it is sometimes possible to improve prediction by measuring conditional intentions.
hanged prior to performance of the
)
These conditional intentions take into account extraneous events, foreseen b iables may appear to have moderating iavior relation if the s:cond of these the investigator, that might produce changes in intentions. Finally, we saw that
'seen in the case of unforeseen extra-long range prediction from intentions will usually be accurate at the aggregate cur, they will tend to change inten-level, even when the measure ofintention does not permit accurate prediction of individual behavior.
een a previous measure ofintention intention behavior relation will tend events but weak in their presence.
neasured after the extraneous event.
PREDICTION OF OUTCOMES 1 into account) a strong intention-l r " moderating variables." Take, for We noted in chapter 3 that investigators are sometimes interested in predictin the behavioral target. An intention outcomes rather than behaviors. The case of predicting weight loss will be dis-change greatly when the person is cussed in chapter 9. Unlike behaviors, outcomes are not completely under a
' >ehavior. In contrast, a person who person % volitional control. For example, whether a woman will have a child in I 11 have developed more realistic ex-the next two years depends not only on her own actions but also on such factors in relatively stable.'
as her husband's behavior, the availability of contraceptives or abortions, and the his point. Imagine a young woman couple's fecundity.
it may appear that the simplest way to predict this outcome would be to
- ounter, appears a bit reserved and i her on several past occasions and measure the woman's intention to have a child in the next two years. Note,how-a front. ' r. B on the other hand is ever, that use of such a measure is based on two assumptions: that the outcome M
4 assume that because both men are in question will result from the performance of certain behaviors and that the in-ntion to ask her out. In their con-tention to produce the outcome willlead to the performance of those behaviors.
tation each man is confronted with When these assumptions are met, then the intention to produce an outcome will may differ greatly. Mr. A expects a predict its occurrence. When they are not met, we cannot expect a strong rela iut his intention. Mr. B, who does tionship.
,is taken aback, perhaps offended, A more promising approach is to identify the behaviors and other factors that avior is consistent with his new in-control the outcome. To predict its occurrence one could then measure the pe It is uarelated to his initialinten-son's intentions to perform each of the identified behaviors and assess the I
ence or absence of the other factors.
, elp preserve a stable intention and te,however, that direct experience nt of more ralistic expectations.
SUMMARY
AND CONCLUSION ifluence the stability ofintentions rate the strength of the intention-In this chapter we have considered some of the issues involved in using in tions to predict behavior. We have noted that although intentions are assumed 3erson's arritude toward the target or au asce ac ah Wm hm too, direct experience may influence tion and behavior depends on two factors: First, the measure of intention has stion eith behavior, to correspond to the behavioral criterion in action, target, context, and time; S
l
i i
k-
]
l l
I l
52 A Theory of Reasoned Action f
I%
l second, a measure ofintention will predict behavior only if the intention does not change before the behavior is observed.These considerations apply whether the criterion is a single action, a choice between multiple alternatives, a behav-ioral category, or an index based on repeated observations, in fact, we saw that l
l each of these criteria can be viewed as consisting of one or more single actions.
l The investigator can ensure high correspondence between intention and be-havior by obtaining an appropriate measure ofintention.The intention's stability, however,is not under his control. We saw that many variables assumed to mod-erate the strength of the intention behavior relation may do so by influencing the stability of the intention. To maxiraize behavioral prediction the inter.tlon should be assessed just prior to the behavior. When this is not feasible,it may sometimes be possible to improve long-range prediction by measuringintentions conditional upon the occurrence of certain events. We also noted that long range prediction is less problematic in the case of aggregate data than it is at the indi.
ei, viduallevel.
Finally, we have considered the prediction of outcomes, as opposed to be-haviors. Although it is possible to measure intentions to achieve the outcome, the predictive validity ofintentions depends on the extent to which they lead to the performance of behaviors that control the outcome.
e 3, Inte; 1
l l
l
(
s
.r
- )
.I 4
'f
. g{ista"N;;o.
UNITED STATES g.
- o NUCLEAR REGULATORY COMMISSION a
y-WASH 1NGTON, D. C. 20555
-f 5
'%.... /
I May 26, 1987 4
MEMORANDUM.FOR:'
John'Milligan l
Technassociates FROM:-
Emile L. Julian
- cting Chief Docketing and-.ServiceLBranch-1 l-
SUBJECT:
SHoncHAM EXHIBITS
-l 1-y Any documents filed on the ' ope.n record in the SkoetE HAM pro-ceeding and made a part'of the official hearing record as an exhibit'is considered exempt from'the provisions'of the United States Copyright Act,-unless it was originally filed under seal-
'with the court expressly because of copyright concerns.
1 All of the documents sent to=TI for processing-' fall.within the
.]
exempt classification.
i l
l:
i l
l
'l i
l
.s
+
,,4
. '9 t
1
,