Research Artick Attitude-Behavior Relations: A Meta-Analysisof Attitudinal Relevance and Topic by Min-Sun Kim, University of Hawaii at Manoa, and John E. Hunter, Michigan State University The difficulty ofjinding a relationship between attitudes and behavior is one of thegreatest controversies i n recent social science research. Thepurpose of this study was to determine whether attitudinal relevance substantially affects the magnitude of the correlation between attitudes a n d behavior, and whether the effects are content-free. Using meta-analysis, we integratedjindings f r o m 138 attitude-behavior correlations with a total sample size of 90,908. The behaviors we studied ranged over 19 different categories and a variety of miscellaneous topics. Our results showed a strong overall attitude-behavior relationship (r = .791 when methodological artifacts were eliminated. Aspredicted, the higher the attitudinal relevance, the stronger the relationship between attitudes and behavior. This effect held true across diverse content domains. Implications f o r communication theoly a n d practice are discussed. The concept of attitude, typically viewed as a stable underlying disposition,' has played a central role in explaining communication phenomena, particularly the effects of persuasive messages. M o s t research in the area Although formal definitions of attitude vary, most contemporary theorists seem to agree that the characteristic attribute of attitude is its dispositional and evaluative nature. For instance, Ajzen (1988) defines an attitude as a disposition to respond favorably o r unfavorably to an object, person, institution, or event. Rokeach (1967, 1968) also defines an attitude as a set of interrelated predispositions to action organized around an object or situation. .. ~~ . ~~ ... ~~~~~~~~~ ~ ~ Min-Sun Kim is an assistant professor in the Department of Speech at the University of Hawaii. John E. Hunter is a professor in the Department of Psychology at Michigan State University. The authors wish to gratefully acknowledge the constructive comments and suggestions o n earlier drafts of this paper provided by Professor Steven R . Wilson and Professor Gerald R. Miller of Michigan State University. Copyright 0 1993Journal of Communication 43(1), Winter. 0021-9916/93/$0.0+.05 101 Journal of Communication, Winter 199.3 of communication and persuasion has centered around attempts to change attitudes toward some object or target, seeking to build theories that accurately explain and predict patterns of complex communication behaviors. Underlying these efforts has been the implicit assumption that behavior toward the object will change automatically with the attitude. Evidence has not always supported this assumption, however, and the difficulty of finding a strong, predictive relationship between attitudes and behavioral tendencies has turned into one of the greatest controversies in the social sciences. Scholars have widely debated the relationship between attitudes and overt behavior (for reviews, see Canary & Seibold, 1984; Eagly & Himmelfarb, 1978; Fishbein bi Ajzen, 1972; Liska, 1975; Miller, 1968; Schuman & Johnson, 1976; Seibold, 1975; Wicker, 1969).2As a result, three basic positions have emerged. The first, argued mostly by the behaviorist camp, posits that since attitudes are cognitive events, they have no consequences for the way people act o r the way they perform those acts. Therefore, attitudes cannot predict behavior (see Bandura, 1969; Blumer, 1955; Corey, 1937; Deutscher, 1966, 1973; Doob, 1947; Green, 1954; LaPiere, 1934; Larson & Sanders, 1975; Merton, 1949). Proponents of this position conceive of verbally expressed attitudes as behavior or as surrogates of behavior, hence attitude-behavior (A-B) relations are not particularly relevant or important. They urge abandoning the general concept of attitude and directly studying overt behavior. For example, Bandura (1969) suggests that we might better treat self-reports simply as another class of behavior rather than as indices of an underlying state endowed with special causal powers. Similarly, Green (1954) and Larson and Sanders (1975) also claim that in studying persuasion, we must be concerned primarily with how people behave, not with their affective disposition. A second group of scholars eschews the extreme behaviorist position and claims that attitude is weakly and inconsistently related to behavior (Bowers, 1968; Carr & Roberts, 1965; Crespi, 1971; DeFleur & Westie, 1963; Ehrlich, 1969; Hyman, 1949; Liska, 1974b; Tittle & Hill, 1967a; Weissberg, 1965; Wicker, 1969, 1971). They declare a necessary inconsistency between verbal-scale scores and other overt actions, pointing to situational or individual factors as the crucial determinants of behavior. Based largely on what Warner and DeFleur (1969) have termed the postulate of contingent consistency, this group argues that the influence of the “other variables” is so overwhelming that we should expect little o r no direct relationship between verbally expressed attitudes and overt behav- ’ Another question of theoretical interest stems from the direction of the A-B relationship. While most A-B studies hold that attitudes cause behavior, some hold that behavior determines attitudes (Bern, 1968) or that the two are mutually reinforcing-that a recursive causal relation exists (Kelman, 1974). Given that simple correlations say little about attitudinal versus behavioral forces, we d o not imply directional causality to findings of relationships. 102 Attitudc-Behavior Relations: A .Meta-Andy.sis ior. Recently, researchers have tried to identify and explain more of these other variables or conditions as determinants of behavior. Some of these have included: (a1 individual characteristics, such as cognitive complexity (e.g., O’Keefe & Shepherd, 19821, self-monitoring (e.g., Snyder, 1982; Snyder & Kendzierski, 1982; Snyder & Swann, 19761, private self-consciousness (e.g., Underwood & Moore, 19811, involvement (e.g., Sivacek & Crano, 1982), and self-awareness (e.g., Pryor, Gibbons, & Wicklund, 1977); (b) attitudinal qualities, such as direct experience with the attitude object (e.g., Borgida & Campbell, 1982; Fazio, Powell, & Herr, 1983; Fazio & Zanna, 1978a, 1978b, 1981; Regan & Fazio, 1977), attitude accessibility (e.g.,Fazio, 1986; Fazio, Chen, McDonel, & Sherman, 1982; Fazio, Sanbonmatsu, Powell, & Kardes, 1986; Fazio & Williams, 19861, temporal stability of attitudes (e.g., Kelly & Mirer, 1974; Schwartz, 1978), and the consistency between affective and cognitive responses (e.g., Norman, 1975); and (c) situational normative factors, such as social constraints (e.g., Frideres, 1971; Warner & DeFleur, 19691, norms (e.g., Bowers, 1968; Fendrich, 19671, reference group norms (e.g., DeFriese & Ford, 19691, and social support (e.g., Bellin & Kriesberg, 1967; Liska, 1974a, 1974b). A third group of researchers has suggested that construct-valid attitudes and corresponding behavioral tendencies are closely related to each other, whatever the causal direction might be. One of the directional assumptions comes from the consistency theory, which claims that tendencies to reach consistency between behavior and attitude will act in both directions (see Cohen, 1964; Festinger, 1957; Insko & Schopler, 1967; Kelman, 1974; Rosenberg, 1965). The most commonly evoked directional position, however, has been a causal link from attitudes to behavior (for empirical results, see Andrews & Kandel, 1979; Bentler & Speckart, 1979, 1981; Kahle & Berman, 1979; Kahle, Klingel, & Kulka, 1981). For example, Kahle and Berman (1979) reported a cross-lagged correlational analysis showing that attitudes have causal priority over behavior. This basic assumption of close linkage has prompted several prominent models explaining the extent to which people act according to their attitudes. In the following section, we discuss this third position in more detail. T h e R e l e v a n c e of R e l e v a n c e While different theories contain different operational forms and conceptions of attitudes, it is not difficult to trace the common thread running through these diverse approaches. We propose that different A-B models can illustrate divergent approaches toward enhancing the conceptual match between attitudinal and behavioral elements. We examine how these explanations and theories strongly suggest that the key to predicting behavior via attitudes lies in increasing the conceptual linkage (match, or relevance) between attitude construct and behavior. Rokeach (1967, 1968) argued that the disappointing empirical findings on A-B relations merely suggest the need for a clearer conceptualization Journal of Communication, Winter 1993 of the attitude concept. To obtain some consistency between attitudes and behavior, he proposed that attitude toward the situation be considered in addition to the traditional measure of attitude toward the object. Rokeach and Kliejunas (1972) showed that the two interacting attitudes reliably predict behavior to some extent. Clearly, these findings show the need to focus on both types of attitudinal elements as predictors of behavior. In their model of reasoned action, Fishbein and Ajzen (1975) posited two factors that are related to an underlying set of salient beliefs determining behavioral intentions: (a) the attitude toward the act (measured by rating the specific act on evaluative semantic-differential scales), and (b) normative beliefs (measured by questions about perceived expectations of other people multiplied by degree of motivation to comply with these expectations). Using empirical weights, Ajzen and Fishbein (1970, 1972, 1980) showed that these two factors, combined in a multiple regression equation, predicted various behavioral intentions, and/or behavior. Expanding on Fishbein and Ajzen’s model, Kreitler and Kreitler (1976) proposed four cognitive orientation (CO) components-goal beliefs, general beliefs, beliefs about norms and rules, and beliefs about self-as crucial predictors of behavior. They found that the degree to which attitude scales represent all C O components is related to the positive relation between attitudes and behavior. In another attempt to increase the construct validity of attitudes, Rosenberg (1956) formulated the “instrumentality-value” model. Here, attitudes were defined as some combination of beliefs about an object, either blocking or facilitating an individual’s attempts to attain valued states. The relationship between attitudes and behavior is influenced by an individual’s perceived expectancy of goal attainment. Behavior with reference to an attitude object may vary as a function of its perceived instrumentality, that is, the action’s perceived effectiveness in bringing about a desired goal. Attitudinal relevance, or the conceptual match between attitudinal and behavioral elements (Ajzen & Fishbein, 1972; Kreitler & Kreitler, 1976; Rokeach, 1968; Rosenberg, 19561, is the central focus of major explanations of A-B relationships. Various models of these relationships are different attempts to enhance attitudinal relevance by positing a combination of multiple attitudinal components such as the four CO components (Kreitler & Kreitler, 19761, attitudes toward situation and object (Rokeach, 19681, attitude toward the act and normative belief (Ajzen & Fishbein, 19721, and instrumentality (Rosenberg, 1956). The theoretical implication is that attitudinal relevance is the crucial factor in understanding the relationship between attitudes and behavioral intention. The expectation is that the population correlation will increase noticeably with increasing attitudinal relevance, and construct-valid attitudes will have directive influence over behavior. The major problem in testing the relationship between attitude and behavior is the selection of attitudinal constructs that are relevant to behav- 104 AttitudeBehavior Relations: A Meta-Analysis ior; the attitude scales must be conceptually relevant to the behavioral components being predicted. Since measures of attitude predict behavior to the extent that they tap pertinent behavioral elements, attempts to change behavior by changing attitude must consider the degree of attitudinal relevance to the behavior that is to be changed. Clearly, if there is no conceptual basis for expecting that a given attitude will be relevant in predicting behavior, the failure to find a relationship between the attitude and the behavior cannot be taken as evidence that attitudes are unrelated to behavior. The problem is to decide how pertinent and relevant the attitudinal construct is in predicting behavior. Specifically, the conceptual validity of the attitudinal predictors must correspond to the behavioral criterion. A similar notion of congruence between attitude and behavior has been somewhat wrongly referred to as “single-act”versus “multiple-act” by Fishbein and Ajzen (1974). This means that traditional attitude-toward-object measures should not be expected to predict an isolated single act, but should predict “multiple-act criteria.” While latent constructs with multiple indicators generally afford a higher possibility of understanding the construct than those with a single indicator, this assumption is misleading because the important matter is not the number of criteria, but the construct validity of the items. For instance, it is quite possible that multipleact criteria may completely depart from construct validity but still yield quite a high correlation between attitudes and behavior, purely due to increased reliability of the measure. Thus, we have to ask how relevant and conceptually valid the attitude construct is in predicting behavior. Given that attitudinal relevance will be an important moderator in A-B relationships, we advance the following hypothesis: H1: The higher the attitudinal relevance, the higher the correlation between attitudes and behavior. A second objective of our research is to test whether the relation between attitudes and behavior is content-free. Are attitudes and behavior related for some content categories (topics) but not others? By classifying studies according to topic, we can estimate a true effect size of the A-B relation within each subset. Specifically, we want to determine whether content categories interact with the attitudinal relevance to affect the degree of A-B relation. Until the 1960s few researchers studied the role of content categories as potential moderators of the A-B relationship. Even then, the studies mainly focused on race relations (Bastide & Berghe, 1957; Fendrich, 1967a; Linn, 1965) and marijuana use (Albrecht, DeFleur, 2% Warner, 1972). However, over the last 20 years researchers have conducted studies across a wide variety of content categories (e.g., voting behavior, delinquency, drugs, alcohol, birth control, environmental protection, consumer behavior, games, blood donation, and religion). 105 Journal o_f’Communication,Winter 199.3 One way to investigate the role of behavioral topic would be to categorize studies under different topics that have been commonly investigated. For instance, past reviewers (e.g., Ajzen & Fishbein, 1977; Schuman & Johnson, 1976; Schwartz, 1978) identified voting as a unique behavior because it is strongly and consistently related to attitudes. Bentler and Speckart (1979) also found that the effects of intention depend on the behavior domain chosen. Although the overall A-B correlation might differ across topics, there is no conceptual basis for assuming that the effect of attitudinal relevance holds true for some topics but not for others. By increasing conceptual validity of a given attitude in predicting behavior, we should be able to increase the relationship between the attitude and the behavior regardless of the topic area investigated. Therefore, we advance the following hypothesis: H2: The effect of attitudinal relevance will hold across all topics In addition to diverse theoretical inquiries into the A-B relationship, there have been persistent attempts to summarize the empirical findings. A number of researchers have compiled reviews since Wicker first summarized many of the A-B studies in 1969. In the following section, we present and discuss some of these review studies, pointing out their weaknesses in integrating findings. Finally, we introduce meta-analysis as a powerful and valid alternative to present review studies. Our literature search yielded 11 review studies that attempted to draw some conclusions about the A-B relationship (see Table 1). We excluded theoretical review articles (e.g., Deutscher, 1966; Eagly & Himmelfarb, 1978; Ehrlich, 1969; Kelman, 1974; Liska, 19741). Vastly different views about the usefulness of attitude constructs in predicting behavior were matched by varied review findings (see last column, Table 1). Initial investigators found low to nonexistent correlations between attitudes and behaviors, and reviewers were quite pessimistic about finding a relation. In his review of the A-B literature, Wicker (1969) claimed, “It is considerably more likely that attitudes will be unrelated o r only slightly related to overt behaviors than that attitudes will be closely related to actions” (p. 65). Similarly, in 1964, Festinger expressed surprise that everyone assumed attitudes caused behavior, yet few researchers had sought evidence for the relationship. He searched the literature for such studies and found only three-one, according to him, “of dubious relevance and one of which required reanalysis of data” (p. 405). A third study was acceptable. He concluded that there was little research evidence of a link between attitudes and behavior. Other reviewers pointed to different factors and conditions to explain the low correlation. Tittle and Hill (1967a) listed 15 A-B studies and concluded that low correlations are in part due to the use of single-item measures (presumably low in reliability). Similarly, Fishbein and Ajzen (1 972) 106 AttitudtLBehuviw Relations: A Meta-An&sis found that the general pattern of results is one of nonsignificance for studies using the single observation of a single act. On the other hand, they found that multiple-act criteria significantly relate to attitudes. Kreitler and Kreitler (1976) found 37 cases of positive relation, 68 cases of “no”o r negative relation, and 15 cases of mixed relation. They concluded that the number of CO components is evidently crucial for demonstrating a positive relation. Ajzen and Fishbein (1977) reported that of 46 “not significant” A-B associations, all had low or partial target or action measurement correspondence. In contrast, of 39 “high”A-B associations, 35 had high target and action measurement correspondence. Others-such as Ajzen (19851, Ajzen and Fishbein (19731, Ryan and Ronfield (19751, and Sheppard, Hartwick, and Warshaw (1988)-included only studies that specifically tested Ajzen and Fishbein’s theory of reasoned action. While their findings concerning the theory were mostly promising, their conclusions did not provide much insight into the general problem of A-B relationships. Thus, even though numerous reviewers (some of whom have examined more than 100 studies) have tried to explain the nature of the A-B relationship, many investigators still argue that the demonstrated relationship is essentially 0. Artifacts of Method Our critique points o u t problems regarding the quality of review procedures and explains why the variations in conclusions drawn by most reviews may have been spurious. As we show below, the standard review practices contain serious methodological artifacts. In general, four broad issues can be identified: examination of previous reviews, sampling procedure, type o f synthesis method, and the treatment of methodological artifacts due to sampling error, error of measurement, and dichotomization of variables. Table 1 summarizes and evaluates the prior reviews under those criteria. Criterion I : Examination ofprevious reviews. The critical use o f previous review studies is important in judging the strengths and weaknesses of earlier work. However, only one of the review studies in Table 1 provides any critique of previous reviews. Criterion 2: Sampling procedure. The results of any integrative review will be affected by the population of primary studies upon which the review is focused, and by the manner in which studies are selected from that population (cf. Jackson, 1980). Only 2 of the 11 review studies reported the sources (such as Social Science Citation Index) or the information retrieval systems used to locate primary studies for possible inclusion. When information o n the sources used for locating studies is not available, it is impossible to ascertain the degree of bias in the review results. Another issue is whether a review analyzed a full set of existing studies on a given topic, o r just a subset. Only 4 of the 11 reviews are comprehensive in their selection of studies by considering the full set of 107 + 0 m Selective Comprehensive Selective Purposive Selective Could not be determined Tittle & Hill ( 1967a) Wicker ( 1969) Fishbein & Ajzen ( 1 972) Ajzen & Fishbein (1 973) Seibold (1 975) Ryan & Bonfield ( 1 975) Author by date No No No No No No Criteria for evaluation Critique of Sampling previous method reviews significance 21 35 Average correlation High/moderate/low Significance 16 6 low, 1 moderate low, 2 moderate, 6 high. Conclusions drawn The average correlation between BI and Ba was -44; multiple correlation of attitude, social influence on BI was .60. Half low, half moderate, one high Processes intervening between intention and behavior tend to reduce the relation. A single observation of a single act results in nonsignificant results, while multiple-act criteria result in significant relation. Coefficient of associa- Correlations are rarely above tion reported by .30 and often 0. investigators Low/moderate/high Synthesizing method 14 47 15 K Table 1 : Summary of Reviews on Attitude-Behavior Relationship w 4 0 - Q Selective Comprehensive Ajzen (1 985) Sheppard et al. (1 988) Yes No and B designate behavioral intention and behavior. bA-B designates attitude-behavior. a BI Note: K = number of studies. Comprehensive Ajzen & Fishbein (1 977) No No Comprehensive Kreitler & Kreitler (1 976) Critique of previous reviews No Sampling method Criteria for evaluation Schuman &Johnson Selective (1 976) Author by date Table 1, Continued 6 87 9 142 117 K The theory of reasoned action can provide accurate prediction of intentions and behaviors that are under one’s volitional control. Low correspondence produces mostly nonsignificant relation while high correspondence results in high relation. 34 positive, 68 no or negative, 15 mixed relations. Results vary from small to moderate. Conclusions drawn Estimation of population 0.53 for BI-B, 0.66 for A-Bb. Significance Low if r < .40, high if r > 40. Positivelnegativel mixed Significance Synthesizing method b 9 F b Journal of Communication, Winter 1993 located studies; 6 of the reviews are explicitly s elective.^ Among those reviewers who tried to be comprehensive in sampling, Wicker (1969) included certain classes of studies whose classification as A-B research is not conceptually justified. Much of the research cited by Wicker (1969) correlates work-related attitudes with job performance, which is usually not under one’s volitional control due to differing abilities, skill levels, resources, and so on. (For a discussion of other problems in Wicker’s review, see Dillehay, 1973.) Criterion 3: Type of synthesis method. The way in which characteristics of the primary studies are represented can substantially affect the results and interpretation of the review. It is fairly common for a reviewer to report findings of primary studies in terms of statistical significance. The significance tests d o not provide information on the magnitude and direction of differences or relationships. Furthermore, the 5% error rate in the significance test is guaranteed only if the null hypothesis is true for the population (i.e., if the population A-B correlation is 0). If the null hypothesis is false for the population, then the error rate can be as high as 95% (cf., Hunter, Schmidt, &Jackson, 1982). Even studies that d o not depend on significance tests (e.g., Ajzen & Fishbein, 1977) frequently adopt an arbitrary threshold (e.g., r = .40) that divides studies into high and low A-B relationships. Criterion 4: Sampling error. Most reviews take all variations across studies at face value and report the range of outcome values. This procedure ignores sampling error. Since most studies are done with small samples (i.e., N < 5001, the sampling error is actually quite large, which leads to capitalization on chance in observed outcome values. Criterion 5:Measurement error. Unreliable measurements attenuate the size of the correlation between attitude and behavior. While several reviews address problems of imperfect measurement (e.g., Fishbein & Ajzen, 1972; Tittle & Hill, 1967a), they fail to take the problem directly into account. Variables in the social sciences often are measured poorly. Thus, we would expect uniformity in the literature only if results were corrected to eliminate error of measurement (cf., Hunter et al., 1982). Poor quality in measures of attitude and behavior may have been responsible for many of the low A-B correlations. Therefore, the lack of concern with establishing the reliability of the independent and dependent measure casts doubts upon the credibility of many of the results drawn by past A-B reviews. Criterion 6: Dichotomization. If either attitude or behavior is measured as a dichotomous variable, the value of the A-B relation (actually a pointbiserial correlation) is artificially reduced. A point-biserial correlation is maximum for a 50/50 split; the greater the departure from 50/50, the The information provided by Ryan and Bonfield (1975) was insufficient for making a judgment on this matter. 110 AttitudeBehauior k'elutions; A Metu-Analysis smaller the correlation. Thus, we need to correct for the departure from the 50/50 split, which makes the point-biserial correlation smaller than it could be. If the dichotomy is an extreme split, the reduction can be quite large. Dichotomization of measurement has been quite frequent, especially in the nieasures of behavior in such areas as blood donation, migration, church attendance, and consumer behavior. None of the past reviewers, however, have taken the problem of dichotomization into account, even though single-item, yes-no-type measures pose one o f the most serious threats to valid inference in studies of A-€3 consistency. As we have shown, all the review studies contain major methodological difficulties or shortcomings. The conflicting conclusions drawn by them may be due to spurious variations caused by the presence of any of a number of different artifacts discussed above. In this study, we attempt to minimize these problems by using meta-analysis as developed by Hunter et al. (1982). In contrast to primary analysis, which uses the responses of individuals as data, meta-analysis uses quantitative studies as the unit of analysis. Once the studies are collected, the results of the studies are converted into correlations. While the correlation coefficient from an individual study can be subject to any number of different artifacts (e.g., sampling error, measurement error, range restrictions, artificial dichotomization, computational errors, typographical errors, etc.), we can correct major sources of error at the level of meta-analysis. We expect that when errors due to sampling, reliability of measures, and dichotomization are removed, there will be a significant and substantial relationship between attitudes and behavior unless attitudinal components severely depart from construct validity. To investigate whether our expectations had any factual support, we conducted a series of meta-analyses of past studies. We integrated the findings of 138 A-B correlations based on a total sample of 90,808 ranging over more than 20 different types of activities. The single most useful source for selecting studies relevant to our analysis was a comprehensive bibliography, Attitudes and Behavior: A n Annotated Bibliography compiled by Canary and Seibold (1984). This annotated bibliography lists 600 articles covering theoretical, methodological, and applied aspects of the A-B relationship. Its bibliographic diversity is evident from the varied sources in which these articles appear, including Southern Speech CommunicationJournal, Journal of Marriage and the Family, Journal of Gerontology, Human Communication Research, Public Opinion Quarterly, Journal of Consumer Research, Journal of Marketing, Journal of Personality and Social Psychology, and other major journals of the American Psychological Association and the American Sociological Association. While this list is evidently comprehensive, the criteria for inclusion were far broader than what we wished to adopt for this study; therefore we selected a subset of the articles, primarily quantitative studies. 111 Journal of Communication, Winter 1993 In addtion, we used several indexes such as Social Science Citation Index,and looked under the subject title “attitudes/behavior.” This was especially helpful in finding studies published after 1984. We also obtained references from articles relevant to the topic, and major reviews dealing with A-B relations in general (e.g, Fishbein & Ajzen, 1974; Kreitler & Kreitler, 1976; Schuman &Johnson, 1976; Wicker, 1969)) and in specific domains (e.g., Ryan & Bonfield, 1975; Sheppard et al., 1988;Tittle & Hill, 1967a) for locating studies. This list may not represent all existing studies; however, given the large data set generated by the literature search, it is very unlikely that a few omitted studies could alter the general conclusions of this study. Once we had gathered our full set of located studies, we carefully examined each study according to the criteria listed below. Meta-Analytic Strategy First, a study had to focus explicitly on behavioral prediction based on one’s attitudes. We excluded studies dealing with changes in attitudes or behavior and studies testing the impact of behavior on attitude prediction (e.g., Bruvold, 1972, 1973). Second, each study had to include data that had not been published previously. Thus, we excluded review articles and reanalyses (e.g., Alwin, 1973; Bagozzi & Burnkrant, 1979). In addition, in cases where studies were published twice with the same data, we selected only one of each (e.g., Smetana & Adler, 1979, 1980; Zunich, 1961, 1962). Third, one or more attitude measures (affective and/or cognitive) and behavior measures had to be measured in the study. Conceptually, the attitude construct has been developed from a unidimensional view of a person’s positive or negative affect toward an object (Allport, 1935) to a tripartite o r tricomponential view involving affect, cognition, and conation (see, Bagozzi & Burnkrant, 1979; Kothandapani, 1971; Ostrom, 1969; Rosenberg & Hovland, 1960; Seibold, 1975). Nevertheless, there are few A-B studies that operationalize attitudes multidimensionally. As a matter of fact, the trend is to conceive of attitudes, intentions, and behaviors sequentially (see Bagozzi, 1981; Fishbein & Ajzen, 1975). Any method of attitude assessment that emphasizes its behavioral aspect introduces operational ambiguities, and thus is likely to blur the issue theoretically and empirically. For our purposes, we considered an attitude to be an evaluative feeling that is evoked by a given object, which involves affect and/or cognition. Consequently, we excluded studies that operationalized attitudes as conation only. When a study used conation as part of the attitudinal elements, we tried to separate responses of a conative nature (behavioral intention) from the attitude scale, leaving only affect and cognition elements (Davidson & Morrison, 1983; McGuinness, Jones, & Cole, 1977; Ostrom, 1969). However, we excluded a study by Goodmonson and Glaudin (1971), because it mixed the cona- 112 AttitudeBehavior Relations: A Meta-Analysis tion measure in a single scale with affective and cognitive components, making it impossible to separate behavioral-intent items from the scale. Fourth, studies had to involve intentional actions consisting of a choice between alternatives, or under one’s volitional control. Theorists studying A-B relations are becoming increasingly interested in the requirement that the behavior of interest be under volitional control. A behavior is completely under a person’s control if the person can decide at will to perform or not perform it (see Ajzen & Madden, 1986; Sheppard et al., 1988). To help predict behavior over which people have imperfect control, several investigators have proposed assessing “facilitating conditions” (Triandis, 1964); “behavioral control” (Ajzen, 1985); and “resources” (Liska, 1984). Recently, Ajzen and his associates (Ajzen, 1985; Ajzen & Madden, 1986; Schifter & Ajzen, 1985) have proposed a “theory of planned behavior,” which extends the theory of reasoned action by including the concept of behavior control. According to the theory of planned behavior, the theory of reasoned action (which relies o n intention as the sole predictor of behavior) is insufficient whenever control over the behavioral goal is incomplete. Researchers in cognitive psychology and artificial intelligence also have studied the factors undermining volitional control over behavior in executing plans. They have proposed assessing “resource limitation” (Wilensky, 1983), “ethical constraints” (Bratman, 19871, and “necessary preconditions” such as cost, risk, ability, ethical legitimacy, skill, etc. (Dyer, 1983). If we focus on behavior over which individuals have only limited control, we are bound to observe some inconsistency between attitudes and behavior. Consequently, this study focuses o n those behavioral domains where most individuals are capable of exercising control over the behavior in question. This is an important criterion that typically has been overlooked in past review studies. Without volitional control, individuals might not be able to perform given behaviors, despite their intentions o r positive dispositions, and the relationship between behavioral intention to overt behavior may be reduced. The amount of volitional control individuals have over their behavior varies along a continuum. Topics dealing with strong emotional aversion rather than judgments, preferences, o r evaluation may show low volitional control; therefore, we excluded the following studies which dealt with strong emotional aversions: “snake phobia” (Bandura, Blanchard, & Ritter, 1969), and “fear of insects,” involving the subject’s overt handling of a large cockroach (Fazio, 1969). In addition, we excluded the following topics because of potentially large differences in individuals’ volitional control: ability to control body weight (Ajzen, 1988>,job performance or job absenteeism (e.g., Ilgen & Hollenback, 1977), reading success (Lewis, 1980), and sitting posture (Mehrabian, 1968). Fifth, we excluded studies that used scales measuring personality traits (e.g., dominance, independence, helpfulness) rather than affect and/or beliefs. Distinguishing between attitudes and personality traits is admit- 113 Journal of Communication, Winter 199.3 tedly difficult. Attitudes are directed at a given object or target (a person, thing, or event); personality traits represent general behavioral tendencies without reference to a specific target, context, or time. (See Ajzen, 1988, for further elaboration of the distinction between traits and attitudes.) Finally, we excluded studies that did not supply sufficient information to allow the computation of a correlation between attitude and at least one of the pertinent dependent variable^.^ For example, Bickman (1972) reported the absolute number of people picking u p litter but did not correlate people's attitude with behavior. Heise (1977) also provided insufficient data to compute an effect size. We also excluded studies that provided correlations between attitude and behavioral intention and behavioral intention and behavior without providing the correlation between attitude and behavior. Studies that met all of the above criteria were selected for meta-analysis. Our final selection included 138 correlations. Studies dealing with multiple topics were analyzed separately (e.g., dating, studying, and exercising-Bentler & Speckart, 1981). We averaged correlations and reported the data as a single study if each correlation was based o n responses to fairly similar objects, such as three brands of fruit drinks (Bonfield, 1974) and two best-selling toothpastes (Ryan, 1978). We coded each study by design features that are potential moderator variables. A moderator variable is a variable that causes differences in the correlation between two other variables. If analysis shows that there is a large corrected standard deviation, then it may be possible to explain the variations across studies by relevant moderators. We grouped the studies into categories on the basis of behavioral domains. The content categories present behavioral topics commonly investigated in A-€3 research. The primary criterion we used to select these categories was whether or not a topic was clearly distinguishable from other topics. The following list of domains has been expanded from the topic areas that Canary and Seibold (1984) used in their comprehensive bibliography: 1. Altruistic behavior: helping behavior, such as volunteering to help blind children, volunteering to be a subject. 2. Consumer behavior: buying behavior, such as buying toothpaste, fruit drinks, prescription drugs. 3 . Deviance: antisocial (delinquent) behavior, such as cheating, buying a term paper, gambling o n campus, stealing books from the library. The attitudinal and normative components of Fishbein and Ajzen's behavioral intention model have been argued by Miniard and Cohen (1981) to be operationally inseparable. Ajzen and Fishbein (1970) proposed that attitudes and subjective norms were related to an underlying set of salient beliefs. Consequently, for those studies testing the theory of reasoned action, we chose multiple correlations of attitude toward the act and normative belief in predicting behavior. 114 AttztudeBehavzor h’elataons A Meta-Ana!yszs 4. Environment: conservation and pollution behaviors, such as littering, water conservation, energy saving. 5. Health care: service utilization, such as dental care; preventive health care behavior, such as getting swine flu vaccination. 6. Groupparticipation: behavior such as joining a union, army reenlistment. 7. Race relations: treatment of minorities, behavior related to minority issues. 8. Religion: religious activities, such as church attendance. 9. Voting: voting practices, such as participation, choice of a candidate. 10. Social activities: visiting exhibitions, watching TV, going to a party. 11. Maternal behavior: breast-feeding. 12. Drug and alcohol use: advocacy and use of drugs and alcohol; consumption, such as smoking marijuana, using minor tranquilizers, drinking hard liquor or other alcoholic beverages. 13. Game behavior: game performance in laboratory setting. 14. Class attendance: attendance at college course. 15. Familyplanning: birth control, such as contraceptive use, having a child. 16. Blood donation: donating blood. 17. Classroom behavior: speech communication performance. 18. Migration: moving, applying for public housing. 19. Verdict: mock court decision. 20. Miscellaneous: any other topics that d o not belong to any of the above categories. We attempted to code studies into three relevance categories (low, moderate, and high), according to the degree of relevance of attitudinal constructs with regard to corresponding behavior measures. Attitudinal relevance is defined as the degree of match between attitudinal and behavioral elements. In this study we operationalized it in terms of the content validity of attitude items with regard to the behavior. The content validity of attitude constructs has been considered a condition of methodological correspondence between attitudinal and behavioral measures (Ajzen & Fishbein, 19771, under various names: “generality equivalence” (Liska, 1974a), “the principle of compatibility” (Ajzen, 1988), and “specificity hypothesis” (Page1 & Davidson, 1984). The general tenet of this hypothesis states that predictor and criterion are defined by four elements: action, the target toward which the action is directed, the context in which the action is directed, and the time at which the action occurs. Of the four entities, only action and the target toward which the action is directed have been used to study attitudinal relevance; virtually no studies have used time and context among their attitudinalhehavioral elements. Consequently, we used only action and target to judge attitudinal relevance. Those studies that match both action 115 Journal of Communication, Winter 1993 and target were classified as high match, studies that had only target match were classified as moderate match, studies that had neither action nor target match were classified as low match.5 However, we found the criterion of entity match inapplicable for studies involving multiple behaviors and/or multiple attitudes. Since it was impossible to judge degree of match using entity match for these studies, we used another approach to determine it. First, studies with a general attitude involving only the target but coupled with behaviors (with regard to the target) across different contexts were coded high match only if the behaviors were representative of the general attitude. A study by Weigel and Newman (19761, for example, correlated scores from an attitude scale measuring concern about the broad category of environmental quality with scores on the comprehensive behavioral index, ranging from petition signing to litter pick-ups. Second, studies involving comprehensive attitudes (using many different attitudinal criteria) and actions performed in different contexts (e.g., Werner, 1978) were coded high match only if the content of the attitude items corresponded with behaviors predicted. The basic purpose of meta-analysis is to obtain correct inferences about population correlations given data on sample correlations. The correlation in each individual study can be subjected to three major sources of error that we can eliminate at the level of meta-analysis: sampling error, error of measurement, and artificial dichotomization of measure. (For a comprehensive discussion of meta-analysis used in this study, see Hunter and Schmidt, 1990; Hunter et al., 1982.) To eliminate the effect of sampling error from a meta-analysis, we must transform the distribution of observed correlations into a distribution of population Correlations. Since sampling error cancels out in the average correlation across studies, our best estimate of the mean population correlation is simply the mean of the sample correlations. However, sampling error adds to the variance of correlations across studies. Thus we must correct the observed variance by subtracting the sampling error variance. The corrected variance is still biased upward because it contains variance due to differences in reliability and dichotomization of measures. Hence, the correlation must be corrected for errors due to the remaining two artifacts.' Since there was a large corrected standard deviation in our meta-analysis, we tried to explain the variations across studies by breaking the studies into measured study features (i.e., relevance and topic). We conducted our meta-analyses using programs developed by Hunter (1990a, 1990b,1990c). The two authors compared classifications of match. The number of disagreements was so small that no formal inter-coder reliabilities were calculated. It is important to keep in mind that even a fully corrected meta-analysis will not correct for all artifacts. Even when sampling error, error of measurement, and artificial dichotomization of a continuous variable are compensated for, there still remains reporting error, bad data, and so on (see Hunter et al., 1982). 116 Attitudt%Behavior Relations: A Meta-Analysis Probing the Attitude-Behavior Link Results are presented as follows: First, the overall correlations of attitude-behavior were identified and corrected for sampling error, measurement error, and dichotomization of variables. Then, in order to test the overall moderating effect of attitudinal relevance (Hypothesis 11, we conducted a series of subgroup analyses. Finally, we conducted another series of subgroup analyses to test the effect of attitudinal relevance across topics, or behavioral domains (Hypothesis 2). Table 2 presents a summary of all the studies used in the present metaanalysis. The list demonstrates the wide variety of studies selected from the viewpoint of topic or behavioral domain. We found pertinent data from 138 correlations, based o n a combined sample of 90,808 participants. We expected to find strong overall support for the general predictive utility of the attitude construct when errors due to sampling, reliability of measurement, and dichotomization were removed. A summary of the overall correlation between attitude and behavior is presented in Table 3 . The weighted mean of the correlation between the attitudes and behavior (corrected for sampling error only) was .47. The standard deviation in this distribution of correlations was .14. The chi-square test indicated that the variance in this distribution of correlations was significantly greater than that expected by sampling error alone: chi-square = 3148.61 (df= 137, p < .OOl). Sampling error was only one source of artifactual variation across the studies. We should eliminate other sources of variance such as dichotomization of measures before we look for moderator variables. Of the 138 studies, 14 studies dichotomized the attitude measures (mean p value = ,371 and 60 studies dichotomized the behavior items (mean p value = .34).’ The more the correlation departs from the 50/50 split, the smaller the correlation is. For instance, in a 32/68 split, the point-biserial correlation is 7% smaller than it would be for a 50/50 split. Correcting for the attenuation due to the dichotomization of variables, the mean correlation between attitudes and behavior was .60 (corrected SD = .20). A chisquare test indicated that the variance in this distribution of correlations was also significantly greater than that expected by sampling error alone: chi-square = 3755.11 (df = 137, p < . O O l ) . Measurement error systematically lowers the correlation between variables. If reliability information is available o n each study, then each correlation is separately corrected for attenuation. Of the 138 studies used in our meta-analysis, only 55 reported reliabilities of attitude items or provided sufficient information to calculate reliabilities. Reliabilities of behavior measures were even more sporadic: 21 out of 138 studies. Since ’ Values of p were available for only 48 of the 60 studies with dichotomized behavior measures. 117 m F + Topic (Category") marijuana (1 2) election (9) voting (9) marijuana (1 2) marijuana (1 2) religiosity (3) marijuana (12) college course (1 4) race (7) marijuana ( 1 2) public housing (20) alcohol (1 2) marijuana (1 2) hard drugs (1 2) dating (10) studying (1 7) exercise (1 0) fruit drinks (2) environment (4) speech communication (1 7) destroy property (3) shoplifting (3) disorderly conduct (3) stealing books (3) underline books (3) fake friendliness (3) getting drunk (3) illegal drinking (3) gambling on campus (3) overcutting class (3) Author Acock & DeFleur (1972) Acock & Scott (1980) Ajzen et al. 1 (1 982) Ajzen et al. 2 (1982) Albrecht et al. (1972) Albrecht et al. (1977) Andrews & Kandel(1979) Babrow & O'Keefe (1 984) Bastide & Berghe (1957) Bearden & Woodside (1 978) Bellin & Kriesberg (1967) Bentler & Speckart 1 (1979) Bentler & Speckart 2 (1979) Bentler & Speckart 3 (1979) Bentler & Speckart 1 (1981) Bentler & Speckart 2 (1981) Bentler & Speckart 3 (1981) Bonfield (1974) Borgida &Campbell (1982) Bostrom (1 970) Bowers 1 (1 968) Bowers 2 (1 968) Bowers 3 (1 968) Bowers 4(1 968) Bowers 5 (1 968) Bowers 6 (1 968) Bowers 7 (1 968) Bowers 8 (1 968) Bowers 9(1 968) Bowers 10 (1 968) Table 2: Summary of Studies on Attitude-Behavior Relationships 5,345 5,342 5,341 5,352 5,348 5,332 5,339 5,267 5,321 5,283 50 202 1,394 130 130 204 244 5,258 253 580 251 58 228 228 228 158 158 158 158 12 N 0.61 0.61 0.39 0.37 0.55 0.86 0.25 0.56 0.23 0.68 0.65 0.35 0.24 0.68 0.22 0.37 0.29 0.54 0.33 0.37 0.61 0.54 0.52 0.50 0.53 0.42 0.31 0.76 0.53 0.39 rob 2 1 3 3 3 3 3 3 2 1 2 3 3 3 3 3 3 3 3 3 3 1 3 3 3 3 3 1 2 3 Match -~ % Fa i L L a Topic (Category") energy ballot (4) open housing (7) lawful acts (3) reclaimed water (4) sex preference (20) civil rights (7) marijuana (1 2) amphetamine (1 2) minor tranquilizer (1 2) beer (1 2) cheating (3) birth control pills (1 5) having a child (15) contraceptive use (15) race (7) cheat in college (3) copy other's paper (3) allow others to copy (3) parental violence (3) spousal violence (3) community water (4) civil rights (7) volunteer (1) game (1 3) game (1 3) race (7) racial relations (7) voting (9) religion (8) voting (9) class attendance (14) Author Bowman & Fishbein (1978) Brannon et al. (1973) Brown (1 974) Bruvold (1 972) Calway-Fagen et al. (1979) Carr & Roberts (1 965) Cooketal. l(l980) Cook et al. 2 (1980) Cook et al. 3 (1980) Cook et al. 4 (1980) Corey (1937) Davidson & Jaccard 1 (1979) Davidson & Jaccard 2 (1 979) Davidson & Morrison (1983) DeFriese & Ford (1 969) DeVries & Ajzen 1 (1 97 1 ) DeVries & Ajzen 2 (1 97 1 ) DeVries & Ajzen 3 (1 97 1 ) Dibble & Straus 1 (1980) Dibble & Straus 2 (1 980) Dillehay et al. (1 969) Ewens & Ehrlich (1972) Fazio & Zanna (1 978a) Fazio & Zanna 1 (1 978b) Fazio & Zanna 2 (1 978b) Fendrich (1967a) Fendrich (1967b) Fishbein et al. (1986) Fishbein & Ajzen (1 974) Fishbein & Coombs (1974) Fredericks & Dossett (1 983) Table 2, Continued 77 640 26 1 99 56 332 349 349 349 349 67 244 244 35 1 262 146 146 146 1,070 2,048 145 83 141 33 43 189 24 66 62 318 234 N 0.84 0.53 0.48 0.22 0.30 0.28 0.69 0.66 0.51 0.70 0.02 0.57 0.54 0.85 0.39 0.37 0.43 0.46 0.28 0.20 0.68 0.32 0.32 0.52 0.59 0.30 0.68 0.74 0.75 0.73 0.23 b 3 3 3 1 1 1 3 3 2 3 1 3 3 3 1 3 3 3 3 3 2 2 3 2 2 2 2 3 3 2 1 Matchb b N 0 N 2,325 204 68 436 25 44 62 110 253 228 236 270 49 270 1,218 39 39 39 474 463 463 463 463 355 94 50 260 34 183 359 Topic (Category") family planning (15) marijuana (12) marijuana (12) marijuana (12) erotica (20) race (7) unionization (6) union (6) game attendance (1 4) army reenlistment (6) army reenlistment(6) birth control pills (15) religiosity (8) blood donation (16) race (7) dishonesty (3) violence (3) drug abuse (12) outgoingness(1 0) Carter candidacy (20) Ford candidacy (20) religion (8) drinking (12) drinking (12) church attendance (8) birth control (1 5) Navy reenlistment (6) race (7) college cheating (3) college cheating (3) Author Freedman et at. (1975) Frideres et al. (1 97 1) Frideres & Warner (1 980) Frideres (1 97 1) Gibbons (1975) Green (1 972) Hamner & Smith (1978) Herman (1973) Holman (1 956) Homet al. (1979) Hom & Hulin (1981) Jaccard et al. 1 (1977) Jaccard et al. 2 (1977) Jaccard et al. 3 (1977) Jackman (1 976) Jones 1 (1980) Jones 2 (1 980) Jones 3 (1 980) Kahle et al. (1981) Kahle & Berman 1 (1979) Kahle & Berman 2 (1979) Kahle & Berman 3 (1979) Kahle & Berman 4 (1979) Kilty (1978) King (1975) Kothandapani (1 97 1) LaRocco (1 983) Linn (1 965) Liska (1 974a) Liska (1978) Table 2, Continued 0.62 0.59 0.54 0.59 0.43 0.48 0.42 0.53 0.41 0.65 0.70 0.65 0.65 0.43 0.13 0.41 0.45 0.46 0.62 0.53 0.57 0.57 0.58 0.53 0.84 0.61 0.16 0.39 0.33 0.31 b 3 3 3 1 3 3 3 2 2 1 3 3 3 3 3 1 1 1 2 3 3 3 3 3 3 1 2 2 3 3 Matchb F N F Topic (Category”) race (7) infant feeding (1 1 ) paper recycling (4) breast feeding (1 1 ) volunteer (1) religiosity (8) swine flu (5) church (8) verdict (1 9) verdict (1 9) blood donation (1 6) nursing (1 1 ) capital punishment (20) class attendance (1 4) taking a loan (2) voting (9) altruism (1) marrow donation (1 6) energy saving (4) legal drinking age (20) comprehensive exam (1 7) verdict (1 9) verdict (1 9) mobility (18) energy (20) political participation (9) political participation (9) drinking (1 2) family planning (1 5) voting (9) race (7) Author Mann (1959) Manstead et al. (1983) McGuinness et al. (1977) Newton & Newton (1 950) Norman (1 975) O’Keefe & Shepherd (1982) Oliver & Berger (1979) Ostrom (1 969) Perry (1 976) Perry & Gillspie (1976) Pomazal & Jaccard (1 976) Potter & Klein (1 957) Prisiin (1 987) Rokeach & Kliejunas (1972) Ryan & Bonfield (1 980) Sample & Warland (1 973) Schwartz (1978) Schwartz & Tessler (1 972) Seligman et al. (1 979) Sivacek & Crano 1 (1982) Sivacek & Crano 2 (1982) Snyder & Kendzierski (1 982) Snyder & Swann (1976) Speare (1 974) Stutzman & Green (1 982) Tittle & Hill (1 967a) Tittle & Hill (1967b) Veevers (1971) Vinokur-Kaplan (1978) Warland & Sample (1 973) Warner & DeFleur (1969) Table 2, Continued N 102 215 132 91 189 313 792 145 66 66 270 25 71 81 93 243 192 195 56 93 96 132 120 700 364 30 1 151 75 239 279 73 1 0.28 0.70 0.28 0.29 0.38 0.62 0.26 0.53 0.66 0.66 0.38 0.70 0.68 0.61 0.32 0.29 0.39 0.38 0.74 0.23 0.60 0.53 0.22 0.30 0.34 0.62 0.50 0.72 0.42 0.26 0.10 b 1 2 2 2 3 3 3 3 2 2 3 2 2 2 2 1 2 3 1 2 2 2 2 1 2 3 3 3 2 1 1 Matchb b N N * 8 = religion 9 = voting 10 = social 1 1 = maternal 12 = drug/alcohol use 13 = game 14 = attendance N 758 113 32 44 28 61 488 152 26 17 82 82 82 82 103 25 1 15 = family planning 16 = blood donation 17 = studying 18 = migration 19 = verdict 20 = miscellaneous Topic (Category") blood donation (1 6) environment (4) dental care (5) environment (4) quarter system (20) birth control (15) abortion (15) church (8) puzzle (1 3) beverage (2) buying container (4) high-phosphate detergent (4) unleaded gas (4) colored paper (4) religiosity (8) blood donation (1 6) Match (of attitudinal relevance to behavior): 1 = low, 2 = moderate, 3 = high 1 =altruism 2 = consumer 3 = deviance 4 = environment 5 = health 6 = group 7 = race " Key to topic categories: Author Warshaw et al. (1986) Weigel et al. (1974) Weigel & Amsterdam (1 976) Weigel & Newman (1976) Weinstein (1 972) Werner & Middlestadt (1 979) Werner (1 978) Wicker (1 97 1) Wilson et al. (1984) Wilson & Dunn (1986) Winters l(1971) Winters 2 (1971) Winters3 (1971) Winters 4 (1 97 1 ) Zanna et al. (1980) Zuckerman & Reis (1978) Table 2, Continued b 3 3 0.54 0.36 0.34 0.33 0.33 3 2 3 1 3 3 2 2 Matchb 2 2 3 3 3 3 0.14 0.60 0.55 0.62 0.55 0.55 0.78 0.46 0.54 0.59 0.31 Attitude-Behavior Relations: A Metu-Analysis Table 3: Overall Effect Sizes of Attitude-Behavior Relationship Artifacts corrected Sampling error only Number of r ( K > Total number of subjects Mean correlation Corrected SD Homogeneity chi-square 138 90,808 ,473 ,141 3,148.6 1 Sampling error & dichotomization Sampling error, dichotomization, & measurement error 138 138 90,808 90,808 ,604 ,200 3.755.1 1 ,79 - Note: df= 137. only a small number of studies reported reliability of the measures, it was impossible to correct each correlation o n an individual basis. On the other hand, 125 out of 138 studies reported the number of attitude items, and all 138 studies reported the number of behavior items. The greater the number of items used, the more reliable the score will tend to be. Consequently, we used the number of items employed to measure attitudes and behavior to estimate the average reliability of the measures. The relationship between the number of items in the measure and its reliability is given by the Spearman-Brown formula. Using this formula, we can calculate the estimated reliabilities of a unit-length measure based on the reliabilities provided. This in turn can be used to estimate the average reliability of measures with the average number of items. Using the mean number of attitude items ( M = 7.16) and the mean number of behavior items (A4= 4.071, we calculated the mean reliabilities for each to be r,, = .84 f o r attitudes, and r,, = .71 for behavior. Using the classic formula for correction for attenuation, we calculated the mean correlation between attitude and behavior and found it to be r,, = .79. The results suggest that a strong relationship between attitude and behavior exists, but that the relationship is substantially attenuated by methodological artifacts. The overall average A-B correlation was .47. When the effects of measurement error and dichotomization of variables were removed, the mean correlation was quite substantial ( r = ,791. Since the correction for sampling error indicated a substantial variation in population correlations across studies, it is reasonable to look for a potential moderator variable to explain the variance. Hypothesis 1 stated that the higher the attitudinal relevance, the higher the correlation between attitudes and behavior would be. To examine the potential moderating effect of attitudinal relevance, we conducted separate meta-analyses o n the studies within each of the three attitudinal relevance subcategories (low match, moderate match, and high match). Hunter et al. (1982) suggested that a moderator variable shows itself in 123 Journal of Communication, Winter 1993 Table 4: Subgroup Analysis: Degree of Attitudinal Relevance as a Moderator Artifacts corrected Low match (df = 22) Number of r ( k ) Total number of subjects Mean correlation Corrected SD Homogeneity chi-square Sampling error only Sampling error & dichotomization Sampling error, dichotomization, & measurement error 23 23 23 6,097 6,097 .31 .15 138.01 107.12 Moderate match (df = 37) Number of r ( k ) 38 Total number of subjects 11,441 Mean correlation .49 .14 Corrected SD Homogeneity chi-square 402.15 High match ( d f = 76) Number of r ( k ) Total number of subjects Mean correlation Corrected SD Homogeneity chi-square 6,097 .26 .13 77 38 38 11,441 11,441 .50 .15 .64 .19 453.44 77 73,270 .40 .19 77 73,270 73,270 .49 .13 .67 .18 2,133.60 2,166.12 .86 .23 two ways: (a) the average correlation will vary from subset to subset, and (b) the corrected variance will be lower in the subsets than for the data as a whole. As shown in Table 4, the results of these subgroup analyses were consistent with the hypothesis, showing that the effect size in each of the three conditions increased as attitudinal relevance increased. When corrected for dichotomization, the average effect size (measured as r ) was .31 ( k = 23; n = 6,097) in the low-match condition, .50 ( k = 38; n = 11,441) in the moderate-match condition, and .67 ( k = 77; n = 73,270) in the high-match condition. Using the estimated average reliabilities for both attitude and behavior measures (raa= .84; r,, = .71), the correlations were corrected for measurement error also, with these results: requaled .40 for low match, .64 for moderate, and .86 for high). The magnitude of the increase in correlations was quite impressive, lending support to the important role of attitudinal relevance in the prediction of behavior. In addition, the true variance in each of the three subgroup estimates was less than the true variance in the overall estimate ( S D = .20); corrected standard deviation was .15 for low match, .15 for moder- 124 Attitude-Behavior Relations: A Meta-Analysis Table 5: Unit-length Reliabilities Across Topic Unit-length reliability of measures Topic" Altruism Deviance Environment Health Group Race Religion Voting Social activity Drug/alcohol Attendance Family planning Blood donation Studying Miscellaneous Overall K Attitude Behavior 3 22 12 2 4 12 8 8 3 19 4 9 4 3 8 .11 .17 .35 .64 .52 .24 .28 .63 .32 .49 .47 .71 .42 .50 .47 .44 .24 .70 .43 .2 1 .46 .32 .5 1 121 .42 .44 .65 Note: K = number of studies. Studies relating to consumer, maternal, game, migration, and verdict behaviors were excluded because they did not report reliabilities for any attitude or behavior measures. ate, and .18 for high). However, the obtained variance was larger than that expected by chance. The chi-square tests indicated that the variance in the distribution of correlation was significantly greater than expected due to sampling error alone (see Table 4). Since there was a large amount of variance across studies, it is still possible to test the second potential moderator to explain that variance. We hypothesized that the effect of attitudinal relevance would apply across all content domains, so we conducted separate meta-analyses on studies within different topic categories. Using the number o f items to estimate the average reliability permitted an accurate correction of correlations overall and for the match subgroup analysis. However, we decided that the same procedure might not yield accurate correction for attenuation, since there was a substantial variation among reliabilities of single item by topic (see Table 5 ) . The variability of reliabilities in both attitude and behavior measure across topics was quite substantial. For attitude reliability, the mean was .42 ( S D = .17); for behavior reliability, the mean was .44 ( S D = .16). In addition, 9 out of the 20 topics had three or fewer studies within a category. Using the number of items to estimate reliabilities works best when there is a sufficient number of studies within each category. While at least one study in most topics provided reliability for the attitude measure, in 5 of the 20 topic areas, not a single study provided reliability for the behav- 125 Journal of Communication, Winter 1993 Table 6: Subgroup Analysis: Topic by Match Match level a Topic High Moderate Low Altruism Consumer Deviance Environment Health Group Race Religion Voting Social Maternal Drug/alcohol use Game Attendance Family planning Blood donation Studying Migration Verdict Miscellaneous .35 (2) .69 (1 8) .57 (7) .33 (1) .85 (2) .71 (2) .60 (7) .47 (5) .46 (3) .31 (2) .40 (3) .55 (1) .25 (2) .24 (7) 5 7 (1) .34 (2) - .65 (16) .73 (2) .88 (6) .46 (3) .68 (1) .64 (2) .39 (1) .41 (3) .43 (2) .49 (2) .67 (1) .46 (3) .92 (1) .65 (3) .50 (3) .56 (3) .61 (1) .72 (3) .14(1) .37 (2) 5 3 (4) .53 (5) Overall .67 (77) 5 0 (38) Total correlation Total K .23 (1) .48 (1) .47 (3) .37 .41 .69 51 .34 .68 .31 .60 .49 .46 .65 .56 .56 .56 .83 .28 .46 .48 53 53 3 3 22 12 2 5 12 8 8 3 3 19 3 4 9 4 3 1 4 10 .31 (23) .66 138 - - Note: Correlations are corrected for sampling error and artificial dichotomization. Numbers in parentheses indicate k, number of studies on which correlation is based. ior measure. Consequently, there was a good possibility that a reliability provided by a particular study within a topic was significantly lower or higher than the actual value. This led to a substantial bias in the reliability calculated with the Spearman-Brown formula using the number of items. Consequently, it was not possible to correct for measurement error across topics. While no correction for measurement error could be made and hence the correlations were lower than the true values, the effect of attitudinal relevance certainly appeared to operate across topics (Table 6). Only 3 of the 20 topics (altruism, health, and voting) did not conform to this pattern. However, this divergence occurred for those topics that had singlestudy cells in at least one of the match categories. Thus the deviation may have been due to sampling error. To summarize, the results from these subgroup analyses were mostly consistent with the prediction that the higher the attitudinal relevance, the stronger the relationship between attitudes and behavior. In addition, the principle of match or relevance held true across a wide variety of topics. 126 Attitude-Behavior Relations. A Meta-Analysis We found that the relationship between attitudes and behavior was strongly influenced by attitudinal relevance, that is, the degree of conceptual match between the attitude and the behavior being predicted. For a high match, the mean correlation was .86. For a medium match, it dropped to .64; and for a low match, it dropped to .40. Could it be that these three correlations between attitudes and behavior arose from chance? In all three match levels, the sample sizes for the average correlation were such that there was virtually no sampling error in the mean correlation ( N = 73,270 for high match, N = 11,441 for medium match, N = 6,097 for low match). Could it be that the relationship was “inconsistent” in the sense that only some correlations were positive while many were zero? The standard deviations in Table 4 are overestimates because of failure to correct for variations in reliability and for unmeasured artifacts. However, with this overestimation in mind, confidence intervals were constructed using these standard deviation estimates. The true intervals were narrower. The 95% confidence interval for the high match was .41 to 1.00, which excluded 0. The 95% confidence interval for the medium match was .27 to 1.00, which excluded 0. The 95% confidence interval for low match was .03 to .77. Although the confidence interval did not get down to 0, a fully normal distribution would. However, the probability of a deviation that extreme is about 1 in 50. Since the standard deviation is known to be an overestimate, the true probability is considerably lower than that. Thus for a high or medium level of relevance, the correlation between attitude and behavior will almost always be above 0. For a low match, there is a less than 1 in 50 chance that there will be a setting with a correlation of 0. This is a critical observation because the meta-analysis controls for the effect of sampling error, and allows us to discover the underlying uniformity in the relationship between attitudes and behavior in terms of the moderator effect of attitudinal relevance. The behaviors we studied ranged over 19 specified categories and a variety of miscellaneous topics. There were too few studies to fully fill the 60 cells for topic by match (20 topics X 3 relevance match levels), so we were unable to compute a quantitative assessment of the extent of variation across topics. Examination of Table 6 shows that for every cell present, the mean correlation (attenuated) was positive. The lowest means were .33 for the high match, .14 for the medium match, and .23 for the low match. However, these low means all occurred for single-study cells where sampling error was maximum. Searching for the lowest number among a set of numbers with high sampling error caused a considerable capitalization on chance. Thus, it was likely that each of the low cells was low because of negative sampling error. If only cells with two or more studies are considered, the lowest cell means are .35 for high match, .37 for the medium match, and .31 for the low match. These means are not corrected for error of measurement: so the true minimum means are still 127 Journal of Communication, Winter 1993 higher. The data suggested that there is no topic for which the correlation between attitude and behavior would be 0. The high-match cells with three or more studies all had means that would correct to about the observed average across all high-match studies-that is, all were in a region near .86. W h y Others Erred Using meta-analysis, we found uniformly positive correlations between attitude and behavior with a very high correlation if the attitude was highly relevant to the behavior. How could so many narrative reviews and other studies come to the conclusion that attitudes only sporadically predict behavior? The answer lies in the known methodological problems of small-sample research in areas without established measurement scales. The meta-analysis showed that past reviews were strongly influenced by the randomness of sampling error, by attenuation due to measurement error, and by attenuation due to artificial dichotomization. Many researchers have based their conclusions on results from some one study that they either personally favor or with which they were closely connected. But isolated studies are subject to massive sampling error and hence massive error in the statistical significance test. Consider the low-match studies. The mean uncorrected correlation (the basis for significance tests) was .26. Fifty percent of the studies were done with a sample size of 102 or less (some much less). For a sample size of 102, a sample correlation must be at least .20 to be significant at the .05 level. Since the standard error of a correlation computed from 102 cases is '10, 27% of such studies would find the correlation not significant. In our analysis, correlations in the low-match condition were not homogeneous; the standard deviation of population correlations was .13. Thus, the standard deviation of sample correlations was .16, and the percentage of studies with nonsignificant results was 36. Further, many studies had fewer than 102 subjects. Thus the error rate for the significance test in early low-match studies was probably about 50%. The sampling error guaranteed that about 50% of the researchers who worked in the area would falsely conclude on the basis of personal experience that attitudes d o not predict behavior. Taking into account the sample sizes of typical social science studies, the interpretation of isolated study results using the significance test is subject to massive error. What about reviews that consider the results of multiple studies? Nearly all reviews done before meta-analysis were narrative reviews in which the reviewer considered each study, one at a time. Thus, reviewers who used the significance test (as most did) carried over the high error rate for individual studies into their conclusions about the set of studies. The full set of low-match studies would show that about 50% of them found a significant correlation. Some reviewers in the A-B area (especially behaviorists) interpreted this kind of finding to mean 128 AttitudeBehavior Relations: A Metu-Analysis that attitudes are unrelated to behavior. That is, they assumed that if half the studies found an effect while the other half did not, the effect was a chance event and there was really no relationship. This reasoning is erroneous because it falsely assumes that the chance base rate is 50%. But the significance test is designed so that the chance base rate is 5% rather than 50%. Thus, if 50% of studies find significance, that is 10 times higher than the chance base rate. Other reviewers made a similar error. They assumed that “significant” equals “related“and “not significant” equals “not related.” Hence, they interpreted the 50% significance rate to mean that attitudes are related to behavior in only 50% of all settings; causing what was termed “inconsistency in the A-B correlation.” This conclusion is also false. Even using the overestimated standard deviation of .19 and assuming a normal distribution of true correlations, the number of settings with no correlation would be fewer than 1 in 50. Finally, our results suggest that a strong relationship exists between attitudes and behavior, but that the relationship is substantially attenuated by methodological artifacts. The weighted mean of correlation between attitudes and behavior (corrected for sampling error only) was .47. When corrected for the artificial dichotomization of variables, the correlation became .60. When the effects of both dichotomization and error o f measurement were removed, the correlation increased to .79. This result showed a systematic downward bias in the average correlation due to methodological artifacts. The implication is that improving measures of attitudes and behavior will undoubtedly reduce, if not eliminate, the problems affecting past research. The extent to which attitudes are predictive of social behavior has been viewed from at least three general viewpoints. The first position begins with the premise that no necessary connection exists at all and, thus, it may be desirable to abandon the attitude concept and other verbal predictors and directly study the overt behavior of interest and the variables that affect that behavior. Despite the vast amount of research that supports this pessimistic position, it is clearly negated by the results of our meta-analyses. We found that the A-R relationship, far from being “essentially zero,” is neither as inconsistent nor as inconclusive as it first appears. The average population correlation between attitudes and behavior was quite substantial ( r = .79, corrected for measurement error and dichotomization). As we have discussed, the correlation between attitude and behavior will almost always be above 0 for a high o r medium match. Even for low-match studies, there is a less than 1 in 50 chance that there would be a setting with a correlation of 0. The second position, the “other variables” approach, posits that attitude is weakly and inconsistently related to behavior due to situational or individual factors. ’This position assumes that the strength of A-B relations is contingent on a variety of factors. For this argument to be true, only some correlations would be positive; many would be 0. In our analysis, calcula- 129 Journal of Communication, Winter 199.3 tions of 95% confidence interval showed that correlations in both high and medium match were uniformly positive. Further, in the high-match condition, the relationship between attitudes and behavior was strong enough ( r = ,841 to “override” the situational factors employed. While investigation into situational and contextual determinants of behavior is an important area of research, the lack of an overarching theoretical framework on contingent consistency hinders our understanding of the A-B relationship. Although a large number of personal and situational variables have been suggested, accumulation of empirical data on those variables has not produced an integrated understanding of A-B relationships. Our meta-analysis suggests that the theoretical utility of relevant attitudes seems to outweigh the effect of intervening variables, and thus the relevant attitudes should be the fundamental condition for the A-B relationship, rather than only one of the myriad of variables that are considered when trying to predict overt behavior. The third position, which claims that attitudinal prediction of behavior depends on the conceptual relevance of the attitude construct, is most consistent with our findings. We found that the population correlation noticeably increased with increasing attitudinal relevance. This evidence supports our claim that the focus of A-B research should be the selection of attitudinal constructs that are relevant with regard to behavior. According to our findings, attitudinal relevance is the crucial factor in understanding the relationship between attitudes and action tendencies. We observed evidence for the importance of attitudinal relevance across more than 20 different types of activities, ranging from very simple strategy choices in laboratory games to actions of considerable personal o r social significance (e.g., having an abortion, smoking marijuana, reenlisting). The Importance of Construct-ValidAttitude To summarize, the results of our meta-analyses challenge both the postulate of contingent consistency and the postulate of no relationship between attitudes and behavior. Neither position adequately describes the ways in which attitudes and actions are linked. Our series of meta-analyses offer clear-cut evidence that implies that construct-valid attitudes have directive influence over behavior. Hence, we can unravel the mystery surrounding prediction and explanation of specific action tendencies by turning our attention to construct-valid attitudes that correspond precisely to the particular action tendency of interest. An additional implication of the meta-analyses concerns the individual’s behavioral control. The strong overall A-B relationship ( r = .79) we observed was due in part to the exclusion of studies involving behaviors with relatively low volitional control. The predictive validity of attitudes partly depends on the degree of volitional control over the behavior in question-meaning that people can easily perform these behaviors if they 130 AttitudeBehauior Relatioras: A Meta-Analysis are so inclined, or refrain from performing them if they decide against it. Recently, many investigators have turned their attention to the question of volitional control (e.g., Ajzen, 1988; Ajzen 6i Madden, 1986; Liska, 1784). The high A-B correlation lends indirect support to this issue. The main implication is that actions that are determined mostly b y factors beyond an individual’s voluntary control fall outside the boundaries established for the A-B problem. Most communication and persuasion research has centered around attempts to change attitudes toward some target. In contrast to those who propose that we study affective relations independently of behavior (Chaffee & Lindner, 1969), we must investigate the relationships between underlying attitudes and behavioral indicators to understand and predict the effects of human communication. Researchers have presumed that changes in attitudes lead to changes in overt behavior. While the centrality of the attitude concept remains, researchers are disenchanted with the usefulness of attitude constructs in predicting behavior (see Miller, 1980). The present meta-analyses clearly imply that knowledge of relevant attitudes has important predictive utility. For communication researchers dealing with political and social policy campaigns-who may be interested in knowing whether their efforts to change attitudes will help them achieve their ultimate goal of changing behavior-the answer, under several constraints, appears to be yes. Specifically, three conditions are necessary to expect that attitudes will be translated into actions. First, we must be sure that attitude scales are conceptually relevant to the behavioral components being predicted. Since measures of attitudes can help predict behavior to the extent that they tap the pertinent behavioral elements, attempts to influence behavior by means of attitude change must also consider the degree of attitudinal relevance to the behavior that is t o be changed. Second, we rnust exercise caution to ensure that the behavior in question is toward the volitional side of the continuum. That is, we must consider (from the subject’s perspective) the circumstantial factors surrounding the predicted behavior, especially since individuals may not be free o r able to enact the predicted behavior. For instance, the relation of attitudes to overt behavior may be reduced if the behavior is not entirely under the person’s volitional control. Thus, individuals may not be able to perform given behaviors, despite their intentions to d o so. Third, we should use proper measures of attitude and behavior. Past studies have rarely reported reliabilities of measures, and most still use single-item dichotomized behavior measures (e.g., to donate blood o r not). It is essential in A-B research to establish and report the reliability of measures. If a behavior o r attitude is measured in a dichotomous manner, a suitable correction should be made. Attitude-behavior relationships have been an interdisciplinary concern. The underlying assumptions and corresponding empirical findings are crucial not only to the theoretical development of human communication 131 Journal of Communication, Winter 1993 studies but also to other fields of applied social science, including consumer behavior, social policy, public campaigns, and others. Insko and Schopler (1967) have suggested the possibility that much evidence showing a close relationship between attitudes and behavioral responses has been obtained but never published because investigators and journal editors have considered such findings “unexciting” and “not worthy of publication.” Our meta-analyses should correct the impression that A-B inconsistency is the more common phenomenon. Providing a critical analysis of the A-B literature makes an important contribution, not only by affirming the utility of attitude constructs in predicting behavior, but in avoiding immense wasted effort in replicating where data already exist to solve the issue. Furthermore, studies that d o not have the characteristics necessary for demonstrating a substantial positive relation between attitudes and behavior should not be taken as evidence that no such relation exists. Ever since Gordon Allport (1935) described the attitude concept as the primary building block in the edifice of social psychology, many researchers have attempted to clarify the A-B relationship. We have tried to thoroughly explore the research literature on A-B relationships to integrate discrepant findings about the strength of such relationships. The results of our meta-analysis lead us to conclude that it is no longer very meaningful to ask “Are attitudes necessary?” Nor does the crucial issue have to d o with searching for more “other variables.” Instead, evidence from the accumulated literature affirms the following position: Relevant attitudes strongly predict volitional behavior. References Acock, A. C., & DeFleur, M. L. (1972). A configurational approach to contingent consistency in the attitude-behavior relationship. American Sociological Review, 3 7, 714-726. Acock, A. C., & Scott, W. (1980). A model for predicting behavior: The effect of attitude and social class o n high and low visibility political participation. Social Psychology Quarterly, 43(1), 59-72. Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In K. Kuhl & J Beckman (Eds.), Action-control: From cognition to behavior (pp. 11-39). Heidelberg, Germany: Springer Publishing Company. Ajzen, I. (1988). Attitudes, personality, and behavior. Milton Keynes: Open University Press, Ajzen, I., & Fishbein, M. (1970). The prediction of behavior from attitudinal and normative variables. Journal of Experimental Social Psychology, 6, 466-487. Ajzen, I . , & Fishbein, M. (1972). Attitudes and normative beliefs as factors influencing behavioral intentions. Journal of Personality and Social Psychology, 21(1), 1-9. Ajzen, I . , & Fishbein, M . (1973). Attitudinal and normative variables as predictors of specific behaviors. Journal of Personality and Socialpsychology, 27, 41-57. 132 Attitude-Behavior Relations: A Meta-Analysis Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, 84, 888-918. Ajzen, I., & Fishbein, M. (1980). Understanding attitudes andpredicting social behavior. Englewood Cliffs, NJ: Prentice-Hall, Inc. Ajzen, I., & Madden, T. J. (1986). Prediction of goal-directed behavior: Attitude, intentions, and perceived behavioral control. Journal of Experimental Social Psychology, 22, 453-474. Ajzen, I., Timko, C . , & White, J. (1982). Self-monitoring and the attitude-behavior relation. Journal of Personality and Social Psychology, 4x31,426-435. Albrecht, S. L., Chadwick, B. A., & Alcorn, D. S. (1977). Religiosity and deviance: Application of an attitude-behavior contingent consistency model. Journalf o r the Scientzfic Study ofReligion, 16(3), 263-274. Albrecht, S., DeFleur, L., & Warner, L. (1972).Attitude-behavior relationships: A reexamination of the postulate of contingent consistency. Paczjic Sociological Review (April), 145-168. Allport, G. (1935). Attitudes. In C . Murchinson (Ed.), A Handbook of Social Psychology (pp. 798-844). Worcester, MA: Clark University Press. Alwin, D. F. (1973).Making inferences from attitude-behavior correlations. Sociomety, 36, 253-278. Andrews, K. H., & Kandel, D. B. (1979).Attitude and behavior: A specification of the contingent consistency hypothesis. American Sociological Review, 44, 298-310. Babrow, A,, & O'Keefe, D. (1984). Construct differentiation as a moderator of attitude-behavior consistency: A failure to confirm. Central States SpeechJournal, 35, 160-165. Bagozzi, R. P. (1981).Attitudes, intentions, and behavior: A test of some key hypotheses Journal of Personality and Social Psychology, 41(4), 607-627. Hagozzi, R. P., & Burnkrant, R. E. (1979). Attitude organization and attitude behavior relationship. Journal OfPersonality and Social Psychology, 3 7, 913-929. Bandura, A. (1969).Principles of behavioral modzfication. New York: Holt, Rinehart, & Winston. Bandura, A., Blanchard, E. R., Ritter, B. (1969). Relative efficiency of desensitization and modeling approaches for inducing behavioral, affective, and attitudinal changes. Journal of Personality and Social Psychology, 13, 173-199. Bastide, R., & Berghe, P. (1957). Stereotypes, norms and interracial behavior in Sao Paulo, Brazil. American Sociological Review, 22, 689-694. Bearden, W. O., & Woodside, A. G. (1978). Situational and extended attitude models as predictors of marijuana intentions and reported behavior. Journal of Social Psychology, 106, 57-67. Bellin, S. S., & Kriesberg, L. (1967). Relationship among attitudes, circumstances, and behavior: The case of applying for public housing. Sociology and Social Research, 51, 453-469. Bem, D. J. (1968). Attitudes as self-descriptions: Another look at the attitude-behavior link. 133 Journal of Comm u n ication, Winter 199.3 In A. G . Greenwald, T. C. Brook, & T. M . Ostrom (Eds.), Psychological foundations of attitudes (pp. 197-215). New York: Academic Press. Bentler, P. M., & Speckart, G. (1979). Models of attitude-behavior relations. Psychological Review, KG( 5), 452-464. Bentler, P. M., & Speckart, G. (1981). Attitudes “cause” behaviors: A structural equation analysis. Journal of Personality and Social Psychology. 40(2), 226-238. Bickman, I.. (1972). Environmental attitudes and actions. Journal of Social Psycholog.y, 87, 32 3-3 24. Blumer, H. (1955). Attitudes and the social act. Social Problems, 3, 59-64. Bonfield, E. H. (1974). Attitude, social influence, personal norm, and intention interactions as related to brand purchase behavior. Journal qfMarketing Research, XI, 379-389. Borgida, E., 8r Campbell, B. (1982). Belief relevance and attitude-behavior consistency: The moderating role of personal experience. Journal ofpersonality and SocialPs.ychology,4x21, 239-247. Bostrorn, R. N.(1970). Affective, cognitive, and behavioral dimensions o f communicative attitudes. Journal of Communication, 20, 359-369. Bowers, W. J. (1968). Normative constraints o n deviant behavior in the college context. Sociometry, 31, 370-385. Bowman, II., & Fishbein, M. (1978). Understanding public reaction to energy proposals: An application of the Fishbein model. Journal ofApplied Social Psychology, 8, 319-340. Brannon, R., Cyphers, G., Hesselbart, S., Keane, R., Schuman, H., Viccaro, T., & Wright, D. (1973). Attitude and action: A field experiment joined to a general population survey. American Sociological Review, 38, 625-636. Bratman, M. E. (1987). Intention, plans, andpractical reason. Cambridge, MA: Harvard University Press. Brown, D. W. (1974). Adolescent attitudes and lawful behavior. Public Opinion Quarterly, 38, 98-106. Bruvold, W. H. (1972). Consistency among attitudes, beliefs, and behavior. Journal of SocialPsychology, 86, 127-134. Bruvold, W. H. (1973). Belief and behavior as determinants of attitude. Journal of Social Psychology, 90, 285-289. Calway-Fagen, N., Wallston, B. S., & Gabel, H. (1979). The relationship between attitudinal and behavioral measures of sex preference. Psychology of Women Quarterly, 4, 274-280. Canary, D. J., & Seibold, D. R. (1984). Attitudes and behavior: A n annotated hihliography. New York: Praeger Publishers. Carr, C., & Roberts, S. (1965). Correlates of civil-rights participation. Journal ofsocial Ps.y chology, 67, 259-267. Chaffee, S., & Lindner, J . (1969). Three processes of value change without behavior change. Journal of Communication, 19, 30-40. Cohen, A. R. (1964). Attitude change and social influence. New York: Basic Books. Cook, M . P., Lounsbury, J. W., & Fontenelle, G. A. (1980). An application of Fishbein and Ajzen’s attitudes-subjective norms model to the study of drug use. Journal ofSocial Psychology, 110, 193-201. 134 Attitudc-Behauior Relations: A Meta-Analysis Corey, S. M. (1937).Professed attitudcs and actual behavior. Journal qf Educational P~sy chology, 28,271-280. Crespi, I . (1971). What kinds of attitude measures are predictive of behavior? Public Opinion Qua rtcrly, 3 S , 3 27-334, rhviclson, A. K., & Jaccarcl,J. J . ( 1979). Variables that moderate the attitude-behavior relation: Results of a longitudinal survey. ,Journalof I’ersonali[y and Social Psycholqy, ?. 7, 1364-1376. Davidson, A. K., 8 Morrison, D. M .(1983). Predicting contraceptive behavior from attitudes: A comparison of within- versus acmss-subjects procedures. Journal of Personality and Social Psycho1og.y. 4 5(5), 997- 1009. DeFleur, M. L., Kr Westie, F. R. (1963). Attitude as a scientific concept. SocialForces, 4.2, 17-31, DeFriese, G . H., & Ford, W. S. (1969). Vcrtnl attitudes, overt acts, and the influence of so cia1 constraint in interracial behavior. Social Problems, 16, 493-504. Deutscher, I. (1966). Words and deeds: Social science and social policy. Social Problems. I-XWinter), 235--254. Deutscher, I. (1973). What u’e.say/uhat U J C do: Sentiments and acts. Glenview, IL: Scott, Foresman and Company. DeVries, D. L., & Ajzen, I. (1971). The relationship of attitudes and normative beliefs t o cheating in college. Jozrrnal qfSocial Rsychology, 8.3, 199-207. Dibhle. V., Kr Straus, M. A. (1980). Some social structure determinants o f inconsistency hetween attitudes and behavior: The case 0 1 family violence. Journal ofMarriage and the Family, 42, 71-80. Dillehay, R. C. (1973). On the irrelevance o f the classical negative evidence concerning the effects o f attitiidcs on behavior. American Psychologist, 28, 887-891. Doob. L. W. (1947). The Ixhavior of attitudes. Psychological Reuiew, 23,667-673 Dyer, M. G. (1983) In-depth understanding; A computer model of integratedprocessiljor narrative comprehension. Cambridge, MA: MIT Press. Eagly, A. H., 81 Hinimelfarb, S . (1978). Attitudes and opinions. Annual Review ofPsycholog . ~ 29, , 517-554. Ehrlich, H. J . (1969). Attitudes, behavior. and the intervening variables. The American Sociologist, 4(1),29-34. Ewens, W. L., & Ehi-lich, €1. I. (1972). Referent-other support and ethnic attitudes as predictors of intergroup behavior. Sociological Quarterly, I 3 , 348-360. Fazio, A . F. (1969). Verbal and overt-behavioral assessment of a specific fear. journal of Consulting and Clinical Psychology, 3.3, 705-709. Fazio, R. H. (1986). How d o attitudes guide behavior? In R. M. Sorrentino & E. T. Higgins (Eds.), 7be handbook of motivation and cognition. Foundations ofsocial behavior New York: Guilford Press. Fazio, R . H., Chen, J., McDonel, E. C., & Sherman, S. J. (1982). Attitude accessibility, attitude-behavior consistency, and the strength of the object-evaluation association. Journal of Experimental Social Psychology, 18, 339-357. Fazio, R. H., Powell, M . C.. & Herr, P. M . (1983). Toward a process model of the attitude-behavior relation: Accessing one’s attitude upon mere observation of the atti tude object. Journal ofPersonality and Social Psychology, 44, 723-735. 135 Journal of Communication, Winter 1 9 3 Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). O n the automatic activation of attitudes. Journal of Personality and Social Psychology, 50(2), 229-238. Fazio, R. H., &Williams, C. J. (1986). Attitude accessibility as a moderator of the attitude perception and attitude-behavior relations: An investigation of the 1984 presidential election. Journal ofPersonality and Social Psychology, 51, 505-514. Fazio, R. H., & Zanna, M. P. (1978a). Attitudinal qualities relating t o the strength of the attitude-behavior relationship. Journal of Experimental Social Psychology, 14, 398-408. Fazio, R. H., & Zanna, M. P. (1978b). O n the predictive validity of attitudes: The roles of direct experience and confidence. Journal of Personality, 46, 228-243. Fazio, R. H., & Zanna, M. P. (1981). Direct experience and attitude-behavior consistency In L. Berkowitz (Ed.), Advances in Experimental Social Psychology, 14, 161-202. Fendrich, J. M. (1967a). Perceived reference group support: Racial attitudes and overt behavior. American Sociological Review, 32, 960-970. Fendrich, J. M. (1967b). A study of the association among verbal attitudes, commitment and overt behavior in different experimental situations. Social Forces, 45, 347-355. Festinger, L. (1957). A theo y of cognitive dissonance. Stanford, CA: Stanford University Press. Festinger, L. (1964). Behavioral support for opinion change. Public Opinion Quarterly, 28, 404-4 17. Fishbein, M., & Ajzen, I. (1972). Attitudes and opinions. Annual Review of Psychology, 23, 487-544. Fishbein, M., & Ajzen, I. (1974). Attitudes towards objects as predictors of single and multiple behavioral criteria. Psychological Review, 81(1), 59-74. Fishbein, M., & Ajzen, I. (1975). BelieJ;attitude, intention, and behavior: An introduction to the0 y and research. Reading, MA: Addison-Wesley. Fishbein, M., & Coombs, F. S. (1974). Basis for decision: An attitudinal analysis of voting behavior. Journal ofApplied Social Psychology, 4, 95-124. Fishbein, M., Middlestadt, S., & Chung, J. (1986). Predicting participation and choice: Firsttime voters in U.S. partisan elections. In s. Kraus & R. M. Perlott (Eds.), Mass media and political thought: An information-processinR approach (pp. 65-82), Berverly Hills, CA: Sage. Fredericks, A , , & Dossett, D. (1983). Attitude-behavior relations: A comparison of the Fishbein-Ajzen and the Bentler-Speckart Models. Journal of Personality and Social Psycholog ~45(3), , 501-512. Freedman, R., Hermalin, A,. & Change, M. (1975). Do statements about desired family size predict fertility? The case of Taiwan, 1967-1970. Demography, 1x31,407-416. Frideres, J. (1971). Situational and personality variables as influencing the relationship between attitudes and overt behavior. Canadian Review of Social Anthropology, 8, 91-105. Frideres, J., & Warner, L. (1980). Attitude-action relationships. Canadian Review of Social Anthropology, 1 7 ( 2 ) , 109-121. Frideres, J., Warner, L . , & Albrecht, S. (1971). The impact of social constraints o n the relationship between attitudes and behavior. Social Forces, 50, 102-112. Gibbons, F. X. (1975). Sexual standards and reactions to pornography: Enhancing behavioral consistency through self-focused attention. Journal of Personality and Social Psychology, 36(9), 976-987. 136 AttitudeBehavior Relations: A Meta-Analysis Goodmonson, C., & Glaudin, V. (1971). The relationship of commitment-free behavior and commitment behavior: A study of attitude toward organ transplantation. Journal of Social Issues, 4, 171-183. Green, B. F. (1954). Attitude measurement. In G. Lindzey (Ed.), Handbook of socialpsychology (Vol. 1). Reading, MA: Addison-Wesley. Green, J . A. (1972). Attitudinal a n d situational determinants of intended behavior towards blacks. Journal ofPersonality and Social Psychology, Z X l l ) , 13-17. Hamner, W. C., & Smith, F. (1978). Work attitudes as predictors of unionization activity Journal ofApplied Psychology, 63(4), 415-421. Heise, D. (1977). Group dynamics and attitude-behavior relations. Sociological Methods G Research, 5(3), 259-288. Herman, J. B. (1973). Are situational contingencies limiting job attitude-job performance relationships? Organizational Behavior and Human Performance, 10, 208-224. Holman, P. (1956). Validation of a n attitude scale as a device for predicting behavior. Journal ofApplied Psychology, 40(5), 347-349, Hom, P., & Hulin, C. (1981). A competitive test of the prediction of reenlistment by several models. Journal ofApplied Psychology, 66(1),23-39. Hom, P., Katerberg, R., & Hulin, C. (1979). Comparative examination of three approaches to the prediction of turnover. Journal ofApplied Psychology, 64(3), 280-290. Hunter, J . E. (1990a). METADICH: Aprogram to do meta-analysis of correlations corrected f o r dichotomization. East Lansing: Michigan State University, Department of Psychology. Hunter, J . E. (1990b). METADIRL: Aprogram to do meta-analysis correcting f o r dichotomization and random error of measurement. East Lansing: Michigan State University, Department of Psychology. Hunter, J . E. ( 1 9 9 0 ~ ) RELMETA: . Aprogram to do meta-analysis of Spearman-Brown reliabilities. East Lansing: Michigan State University, Department of Psychology. Hunter, J . E., & Schmidt, F. L. (1990). Methods of meta-analysis. Newbury Park, CA: Sage Hunter, J. E., Schmidt, F. L., &Jackson, G. B. (1982). Meta-analysis: Cumulating research @dings across studies. Beverly Hills, CA: Sage. Hyman, H. H. (1949). Inconsistencies as a problem in attitude measurement. Journal of Social Issues, 5(Summer), 38-42. Ilgen, D. R., & Hollenback, J. H. (1977). The role of job satisfaction in absence behavior. Organizational Behavior and Human Performance, 19, 148-161. Insko, C. A., & Schopler, J. (1967). Triadic consistency: A statement of affective-cognitiveconative consistency. Psychology Review, 74, 361-376. Jaccard, J., King, G. W., & Pomazai, R. (1977). Attitudes and behavior: An analysis of specificity of attitudinal predictors. Human Relations, 30, 817-824. Jackman, M. R. (1976). The relation between verbal attitude and overt behavior: A public opinion application. Social Forces, 54,646-668. Jackson, G. B. (1980). Methods for integrative reviews. Reviews of Educational Research, ~50(3),438-460. Jones, J. W. (1980). Attitudinal correlates of employees' deviance: Theft, alcohol use, and nonprescribed drug use. Psychological Reports, 47, 71-77. 137 rournal of Communication, Winter 199.3 Kahle, L. R., & Berman, J . J. (1979). Attitudes cause behaviors: A cross-lagged panel analysis. Journal of Personality and Social Psychohgy, .? X 3 ) , 315-321. Kahle, L. R., Klingel, D. M., & Kulka, R. A. (1981). A longitudinal study of adolescents’ attitude-behavior consistency. Public Opinion Quarterly, 45, 402-414. Kelly, S., & Mirer, R. W. (1974). The simple act of voting. ilmerican Political Science Review, 68, 572-591. Kelman, H. C . (1974). Attitudes are alive and well and gainfully employed in the sphere o f action. American Psychologist, 29, 310-324. Kilty, K. M. (1978). Attitudinal and normative variables as predictors of drinking behavior Journal ofstudies on Alcohol, 39, 1178-1194. King, G. W. (1975). An analysis of attitudinal and normative variables as predictors of in tentions and behavior. Speech Monographs, 42, 237-244. Kothandapani, V. (1971). Validation of feeling, belief, and intention to act as three components of attitude and their contribution to prediction of contraceptive behavior. Journal of Personality and Social Psychology, lR3), 321-333. Kreitler, H., & Kreitler, S. (1976). Cognitive orientation and behavior. New York: Springer Publishing Company. LaPiere, R. T. (1934). “Attitude vs Actions.” Social Forces, 13, 230-237 LaRocco, J . M. (1983). Job attitudes, intentions, and turnover: An analysis of effects using latent variables. Human Relations, 36(9),813-826. Larson, C., tk Sanders, R. (1975). Faith, mystery, and data: An analysis of “scientific” studies of persuasion. Quarterly Journal of Speech, 61, 178-194. Lewis, J. (1980). The relationship between attitudes toward reading and reading success Educational and Psychological Measurement, 40, 261-262. Linn, L. (1965). Verbal attitudes and overt behavior: A study of racial discrimination. Social Forces, 4.3(3), 353-364. Liska, A. E. (1974a). Attitude-behavior consistency as a function of generality equivalence between attitude and behavior objects. Journal of Psychology, 86, 217-228. Liska, A. E. (1974b). Emergent issues in the attitude-behavior consistency controversy. American Sociological Rez/iew, 39, 261-272. Liska, A . E. (Ed.). (1975). The consistency controz)ers.y:Readings on the impact of attitude on behavior. New York: Academic Press. Iiska, A. E. (1978). Deviant involvement, associations and attitudes: Specifying the underlying causal structures. Sociology and Social Research, 6.?, 73-89. Liska, A. E. (1984). A critical examination of the causal structure o f the Fishbein/Ajzen attitude-behavior model. Social Psychology Quarterly, 4 X l), 61-74. Mann, J. H . (1959). The relationship between cognitive, affective, and behavioral aspects of racial prejudice. Journal of Social Psychology, 49, 223-228. Manstead, A . S., Proffitt, C., tk Smart, J. L. (1983). Predicting and understanding mothers’ inpdnt-feeding intentions and behavior: Testing the theory of reasoned action. Journal of Personality and Social Psychology, 44(4),657-671. McGuinness, J., Jones, A. I>., & Cole, S. G . (1977). Attitudinal correlates of recycling behavior. Journal ofApplicd Psychology, 6X4), 376-384. Mehrabian, A. (1968). Relationship of attitude to seating posture, orientation, and distance. Journal of Personality and Social Psychology, 10, 26-30. Merton, R. K. (1949). Discrimination and the American creed. In R. M. MacIver (Ed.), Dis- 138 AttitudcLBehatdor Relations: A Meta-Ana@is crimination and national welfare (pp. 99-126). New York: Institute for Religious and Social Studies. Miller, G. It. (1968). Communication and persuasion research: Current problems and prospects. &zuzrterly.rournal of Speech. 54, 268-276. Miller, G. R. (1980). Afterword. In D. P. Cushnian Kr R. D. McPhee (Eds.), Messagc-attitudebehauior relationship: Theoy , mcJthodolog,y,and application New York: Academic Press. Miniard, P. W., Sr Cohen, J. B. (1981). An examination o f the Fishbein-Ajzen hehavioral in tentions model’s concepts and measures. Journal gfExperimentu1 Social Psycholog?;, 17, 309-399. Newton, N. R., & Newton, M. (1950). Relationship o f ability t o breast feed and maternal attitudes toward I)reast feeding. Pediatrics, 5, 869-875. Norman, R. (1975). Affective-cognitive consistency, attitudes, and behavior. Journal of Personality and Social Psychology, 32, 83-91, O’Keefe, I>. J . , 8r Shepherd, G . J . (1982). Interpersonal construct differentiation, attitudinal confidence, and the attitude-behavior relationship. Central States Speech Jour?zal,3.1. 4 16-423, Oliver. K. L., 8r BerSer, P. K. (1979). A path analysis of preventive health care decision models. Journal of C‘onszimerResearch, 6: 113-122. Ostrom, T. M. ( 1969). The relationship between the affective, behavioral, and cognitive components o f attitude. Journal ofllxperimental Social Psychology, 5,12-30. Kr Ihvidson, A. R. (1984). A comparison of three social-psychological models Pagel, M. I]., of attitude and Ixhavioral plan: Prediction o f contraceptive behavior. Journal ofpersonality and Social~.s.sycholog,y,47(3),517-533. Perry, R. W. (1976). The effects of apparent guilt and prior attitudes on the verdicts o f s t u dent juries. Eirropean Journal ofSocia1 I-’sychologly. 6(1),115-1 18. Perry, R. W., 8; Gillespie, 11. F. (1976). An analysis o f intervening variables in the attitude-behavior relationship. Joz*riral oJ’SociulPs-ychology,98, 287-288. I’omazal, R . J , , & Jaccard, J . .J. (1976). An informational approach to altruistic behavior. ,/ournu1ofPersonulity and Social Psycholog-v,-?-3(3),317-326. Potter, €1. W., & Klein, €3. R. (1057). On nursing lxhavior. P.sychintly, 20. 39-46. Prislin, R. (1987). Attitude-behavior relationship: Attitude relevance and Iiehavior relevance. European jou rnal qf Social t’syc h olog.y, 1 7, 483-485. Pryor, J . B., Gibbons, F., 8r Wicklund, R. A . (1077). Self-focused attention and self-report validity. Journul ofPersonalif-y,45, 513-527. Regan 11. T., Kr Fazio, R. H. (1977). On the consistency between attitudes and behavior: Look t o the method of attitude formation. ,/ozir7aal qfExperimenta1 Social Psychology. 1.3, 28-45. Rokeach, M. (1967). Attitude change and behavioral change. Public Opinion Quurtcrly, 30, 529-550. Rokeach, M . (1968) Reliqfi, attitiida, and ualites. San Francisco: Jossey-Bass Rokeach, M., Kr Kliejunas, I-’. ( 1972). Behavior as a function of attitude-toward object and attitude-toward-situation. Journal oJI’ersoriulity and Social Ps-vchology,2 3 2 1 , 194-201 Rosenberg, M. J . ( 1956). Cognitive structure and attitudinal affect. Journal ofAbnormal U T ZSocial ~ Psychologj: 53, 367-372. 139 Journal of Communication, Winter 199.3 Rosenberg, M. J. (1965). When dissonance fails: On eliminating evaluation apprehension from attitude measurement. Journal of Personality and Social Psychology, 1, 28-42. Rosenberg, M. J., & Hovland, C. I. (1960). Cognitive, affective, and behavioral components of attitudes. In C. I. Hovland & M. J. Rosenberg (Eds.), Attitude organization and change (pp. 1-14). New Haven, CT: Yale University Press. Ryan, M. J. (1978). An examination of an alternative form of the Behavioral Intention Model’s normative component. In H. K. Hunt (Ed.), Advances in consumer research 5 (pp. 283-389). Provo, UT: The Association for Consumer Research. Ryan, M. J., & Bonfield, E. (1975). The Fishbein extended model and consumer behavior Journal of ConsumerResearch, 2, 118-136. Sample, J., & Warland, R. (1973). Attitudes and prediction of behavior. Social Forces, 51, 292-304. Schifter, D. B., & Ajzen, I. (1985). Intention, perceived control, and weight loss: An application of the theory of planned behavior. Journal of Personality and Social Psychology, 49, 843-85 1. Schuman, H., &Johnson, M. P. (1976). Attitudes and behavior. In A. Inkeles, J. Coleman, & N. Smelser (Eds.), Annual Review of Sociology, 2, 161-207. Schwartz, S. H. (1978). Temporal instability as a moderator of the attitude-behavior relationship. Journal of Personality and Social Psychology, 36, 715-724. Schwartz, S. H., & Tessler, R. C. (1972). A test of a model for reducing measured attitude-behavior discrepancies. Journal of Personality and Social Psychology, 24, 225-236. Seibold, D. R . (1975). Communication research and the attitude-verbal report-overt behavior relationship: A critique and theoretical reformulations. Human Communication Research, 3-32. al), Seligman, C., Kriss, M., Darley, M., Fazio, R. H., Becker, L. J., & Pryor, J . B. (1979). Predicting summer energy consumption from homeowners’ attitudes. Journal ofApplied Social Psychology, g l ) , 70-90. Sheppard, B. H., Hartwick, J., & Warshaw, P. R. (1988). The theory of reasoned action: A meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research, 15, 325-343. Sivacek, J., & Crano, W. D. (1982). Vested interest as a moderator of attitude-behavior consistency. Journal of Personality and Social Psychology, 43(2), 210-221. Smetana, J. G., & Adler, N. E. (1979). Decision-making regarding abortion: A value x e x pectancy analysis. Journal of Population, 2, 338-357. Smetana, J. G., & Adler, N. E. (1980). Fishbein’s value xexpectancy model: An examination of some assumptions. Personality and Social Psychology Bulletin, 6(1), 89-96. Snyder, M. (1982). When believing means doing: Creating links between attitudes and behavior. In M. P. Zanna, E. T. Higgins, & C. P. Herman (Eds.), Consistency in social behavior: The Ontario Symposium, Vol. 2 (pp. 105-130). Hillsdale, NJ: Lawrence Erlbaum Associates. Snyder, M., & Kendzierski, D. (1982). Acting on one’s attitudes: Procedures for linking attitude and behavior. Journal of Experimental Social Psychology, 18, 165-183. Snyder, M., & Swann, W. B. (1976). When actions reflect attitudes: The politics of impression management. Journal of Personality and Social Psychology, 34(5), 1034-1042. 140 Attitude-Behavior Relations: A Meta-Analysis Speare, A. (1974). Residential satisfaction as a n intervening variable in residential mobility. Demography, 11(2), 173-188. Stutzman, T. M., & Green, S. €3. (1982). Factors affecting energy consumption: Two field tests of the Fishbein-Ajzen Model. Journal of Social Psychology, 11 7, 183-201. Tittle, C. R., 81 Hill, R. J. (1967a). Attitude measurement and prediction of behavior: An evaluation of conditions and measurement techniques. Sociomety, 30, 199-213. Tittle, C. R., & Hill, R. J . (1967b). The accuracy of self-reported data and prediction of political activity. Public Opinion Quarterly, 31, 103-106. Triandis, H. C. (1964). Exploratory factor analyses of the behavioral component of social attitudes. Journal of Abnormal and Social Psychology, 68, 420-430. Underwood, B., & Moore, B. S. (1981). Sources of behavioral consistency. Journal of Personality and Social Psychology, 40, 780-785. Veevers, J. E. (1971). Drinking attitudes and drinking behavior: An exploratory study. Journal of Social Psychology, 85, 103-109. Vinokur-Kaplan, 1). (1978). To have o r not t o have another child: Family planning attitudes, intentions, a n d behavior. Journal of Applied Social Psychology, Cyl), 29-46. Warland, R. H., & Sample, J. (1973). Response certainty as a moderator variable in attitude measurement. Rural Sociology, 38, 174-186. Warner, L., & DeFleur, M. L. (1969). Attitude as a n interactional concept: Social constraint and social distance as intervening variables between attitudes and action. American Sociological Review. 34(2), 153-169. Warshaw, P. R., Calantone, R., &Joyce, M. (1986). A field application o f t h e Fishbein and Ajzen Intention Model. Journal ofsocial Psychology, 126(1), 135-136. Weigel, R. H., & Amsterdam, J. T. (1976). The effect of behavior relevant information o n attitude-behavior consistency. Journal of Social Psychology, 98, 247-251. Weigel, R. H., & Newman, L. S. (1976). Increasing attitude behavior correspondence b y broadening the scope of the behavioral measure. Journal of Personality and Social Psychology, 33, 793-802. Weigel, R. H., Vernon, D., & Tognacci, L. N. (1974). Specificity of the attitude a s a determinant of attitude-behavior congruence. Journal of Personality and Social Psychology, .30(6), 724-728. Weinstein, A. C . (1972). Predicting behavior from attitudes. Public Opinion Quarterly, .?6, 355-360. Weissberg, N. (1965). O n DeFleur and Westie’s “Attitudes as a scientific concept.” Social Forces, 43, 422-425. Werner, P. D. (1978). Personality and attitude-activism correspondence. Journal of Personality and Social Psycholog,y,36(12), 1375-1390. Werner, P. D., & Middlestadt, S. E. (1979). Factors in the use of oral contraceptives by young women. ,JournalofApplied Social Psychology, %6>,537-547. Wicker, A. W. (1969). Attitudes vs. actions: The relationship of verbal and overt behavioral responses to attitude objects. Journal of Social Issues, 25, 41-78. Wicker, A. W. (1971). An examination of the “other variables” explanation of attitude-behavior inconsistency. Journal of Personality and Socialpsychology, IXl), 18-30. Wilensky, R. (1983). Planning and understanding: A computational approach to human reason. Reading, MA: Addison-Wesley. 141 Journal of Communication, Winter 199.3 Wilson, T. D., & Dunn, D. S. (1986). Effects of introspection o n attitude-behavior consistency: Analyzing reasons versus focusing o n feelings. Journal of Experimental Social Psychology, 22, 249-263. Wilson, T. D., Dunn, D. S., Bybee, J. A., Hyman, D. B., & Rotonde, J. A. (1984). Effects of analyzing reasons o n attitude-behavior consistency. Journal of Personality and Social Psychology, 47, 5-16. Winters, L. C. (1971). Measuring attitudes and behavior toward ecologically significant products. Psychological Reports, 29, 893-894. Zanna, M. P., Olson, J. M., & Fazio, R. H. (1980). Attitude-behavior consistency: An individual difference perspective. Journal of Personality and Social Psychology, 38, 432-440. Zuckerman, M., & Keis, H. T. (1978). Comparison of three models for predicting altruistic behavior. Journal of Personality and Social Ps.ychology, 36, 498-510. Zunich, M. (1961). A study of relationships between child rearing attitudes and maternal behavior. Journul uf Experimental Educalion, . 3 @ 2 ) , 231-241. Zunich, M. (1962). Relationship between maternal behavior and attitudes toward children. Journal of Genetic Psychology, 100, 155-165. 142
© Copyright 2026 Paperzz