James C. McCroskey and Samuel V. O. Prichard

The principle of "selective-exposure" can be stated simply: we select that which we like, or, more formally, "people generally tend to expose themselves more readily and more often to messages expressing views compatible with their own attitudes than to messages espousing incompatible views." The following article investigates the broadcast of the 1966 Presidential "State of the Union" Address in light of this principle.

A frequent explanation for the failure of persuasive communications to produce significant attitude change has been that messages conveyed by the mass media permit the operation of the selective-exposure phenomenon. As Lipset et al.1 have observed, people generally tend to expose themselves more readily and more often to communicative messages expressing views compatible with their own attitudes than to messages espousing incompatible views. Formulators of homeostatic theories of attitude and attitude change have made similar observations.2

Numerous studies have provided support for the selective-exposure hypothesis.3 A recent study by Diab, for example, found that Arab students generally tend to listen to radio stations and read newspapers expressing views toward Arab unity compatible with their own.4 In another recent study McGinnies and Rosenbaum found that female students at the University of Maryland who chose to expose themselves to a broadcast of a Lyndon B. Johnson foreign policy speech were significantly more favorable to the President's Viet Nam policy than those who did not attend the broadcast. They did not, however, find a similar difference for male students.5

McGinnies and Rosenbaum pointed to a problem present in their study: namely, the range of attitudes on the part of their male subjects toward the Viet Nam policy was so narrow that there was little opportunity for selective-exposure to function. Therefore, they proposed that "selective-exposure hypothesis should be examined over a wide range of initial attitudes on a salient issue."6 This statement suggested that a study of the selective-exposure hypothesis should concern itself with a single salient issue toward which people hold widely varying attitudes. An examination of the literature relevant to the selective-exposure hypothesis indicates that this is precisely what most researchers have attempted.

An apparent presumption of these studies is that one single attitude dominates the behavioral choices of people when confronted with the option of exposing or not exposing themselves to a given communicative stimulus. This presumes very high correlations between attitudes and behavioral choices. Observed correlations of this magnitude are rare indeed in psychological literature. Thus we suspected that attitudes other than the single attitude reported in each of these studies were functioning in the behavioral choices reported.

Specifically, we suspected that attitude toward the source of a communication as well as attitudes toward issues other than the central issue of a previously announced communication might affect people's decisions to expose or not to expose themselves to a given communicative stimulus, or message. A very large body of research has accumulated which points to the major effect source credibility has on the degree of persuasive affect of a given communicative stimulus.7 Therefore, we hypothesized that people who choose to expose themselves to a communicator's message have more favorable attitudes toward the communicator than those who choose not to expose themselves. A common theory holds that people tend to have groups of compatible attitudes. Thus we hypothesized that on issues other than the central issue of a previously announced communication people choosing exposure would hold more favorable attitudes than people avoiding exposure. For example, our hypothesis would suggest that in the McGinnies and Rosenbaum study cited above the subjects who listened to Lyndon Johnson's foreign policy speech were probably more favorable toward his domestic policies than the subjects who chose not to expose themselves to Johnson's address.


This study took advantage of a major, well-publicized communicative event, the televised presentation of President Lyndon B. Johnson's "State of the Union" address in January, 1966. The day prior to the presentation of the speech, six-scale semantic differentials on twelve substantive issues and on the "authoritativeness" and "character" of President Johnson were administered to 160 students enrolled in a basic speech course at The Pennsylvania State University. (Since the study was conducted the first week of the term, we can assume the students' training in Speech did not affect their behavior.) All measures had been extensively pretested and had been found to be single-factored and reliable.8 The day following the speech the same measures were administered to the subjects. In addition, each subject was asked to indicate whether he had seen and heard the speech, had read the speech and/or heard about its content, or had no exposure to the speech. Each semantic differential, consisting of six seven-step scales, permitted a score range of from 6 (most unfavorable toward the concept) to 42 (most favorable toward the concept). Neutrality was evidenced by a score of 24. So few subjects (five) were exposed to the speech "indirectly" that they were disregarded in the analysis of the data. The subjects were distributed among four cells: one cell each of males and females who heard the speech, and one cell of each of males and females who did not hear the speech. Because of the requirements of available computer programs, the four cells were balanced by random exclusion of subjects. Thus, the 136 subjects involved in the final analysis were distributed 34 to a cell. Although the data for male and female subjects were analyzed separately, the results reported below will disregard the sex factor. No significant differences attributable to sex were observed relevant to the selective-exposure hypotheses.


As shown in Table I, the initial attitudes of the subjects who chose to listen to the presidential address differed significantly from those who chose not to listen only on the concept of "U.S. Policy in South Viet Nam." The traditional selective-exposure hypothesis was supported by this difference, with the subjects who chose to listen to the presidential address significantly more favorable (F = 4.72, p < .05) to the presidential policy. There was a comparatively wide range of initial attitudes on all 14 concepts; but, on the other thirteen concepts no significant differences between exposed and unexposed subjects appeared.

No support was provided by the data for our hypothesis that attitude toward the source of communication functions as a causal factor in the selective-exposure phenomenon. No significant differences on either the authoritativeness or character dimension of source credibility were observed between the subjects who chose to be exposed to the communication and those who chose not to be exposed. In addition, the non-significant differences observed were in the direction opposite to that predicted by this hypothesis.

Our second hypothesis was also unconfirmed. There were no significant differences in initial attitudes between the two groups on any of the eleven remaining concepts. It should be noted, however, that on nine of these eleven concepts the observed non-significant differences in initial attitudes between exposed and unexposed subjects were in the direction of the presumed attitude of President Johnson, and distribution that would be significant beyond the .05 level by the non-parametric sign test.


Mean Attitudes and range of Attitudes for Exposed and Unexposed Subjects

Mean Initial Attitudes

Range of Initial Attitudes

Covariance adjusted Postcommunication Mean Attitudes
Concept Exposed Unexposed Exposed Unexposed
U.S. Policy in South Viet Nama, c 30.8 28.8 6-42 31.5 29.1
Domestic Peace Corps 38.4 36.8 18-42 37.1 37.4
Rent Subsidies for the Poorb 29.8 29.5 6-42 31.0 29.2
U.S. Peace Offensiveb 33.5 32.0 9-42 34.2 32.3
Increased Military Spending 30.0 29.4 6-42 31.0 29.7
Russian Foreign Policy 21.0 21.4 6-42 21.5 21.0
New York Transit Strike 15.4 15.9 6-42 15.8 15.7
Draft Defer. for College Studentsb 36.5 37.1 8-42 37.2 35.5
Red Chinese Foreign Policy 15.5 16.5 6-42 16.7 17.8
Additional Federal Civil Rights Legislationd 32.7 32.1 9-42 34.2 31.0
U.S. Interv. in the Dominican Rep. 25.0 25.6 6-42 26.2 25.1
War on Poverityb 34.5 35.5 6-42 36.6 35.2
Character 29.7 29.8 9-40 30.9 30.4
Authoritativeness 33.4 34.2 9-42 34.5 34.0

aInitial means significantly different, p < .05.

bAdjusted postcommunication means significantly different, p  < .05.

cAdjusted postcommunication means significantly different, p  < .01.

dAdjusted postcommunication means significantly different, p  < .001.

Independent, correlated t-tests indicated that the unexposed subjects' attitudes did not change significantly on any of the twelve concepts or the two dimensions of source credibility.9 Analysis of covariance indicated, however, that the exposed subjects' attitudes on six concepts were significantly modified by the speech. (See Table I.) With initial attitude held constant by analysis of covariance, it was observed that exposure to the speech produced significantly more favorable attitudes (p < .05) toward "U.S. Policy in South Viet Nam," "U.S. Peace Offensive," "Draft Deferments for College Students," "War on Poverty," "Rent Subsidies for the Poor," and "Additional Federal Civil Rights Legislation." In short, the speech was highly successful in modifying relevant attitudes. It should be noted, however, that while the subjects became more favorable toward "U.S. Policy in South Viet Nam" they also increased their favorability toward "Draft Deferments for College Students." The reason for this is not clear. The presidential address made no mention of draft deferments. The subjects could have changed on this concept because they were reassured by the speech that the Viet Nam war could be won without drafting college students. They could also have shifted because they perceived the Viet Nam war to be more personally threatening after hearing the speech. The available data do not enable us to determine why there was attitude change on this concept. It is clear, however, that the speech produced a significant attitude change on a highly salient concept even though there was no direct reference to the concept included in the speech.


It is most likely that prior to exposure the subjects in this study expected the issue of Viet Nam policy to be the most important issue to be discussed by President Johnson in his address. Considerable advance publicity of the schedule speech was available on television, radio, and in the campus newspaper. Views expressed by most reporters pointed to Johnson's concern about his lagging support for his Viet Nam policy, and most suggested that this would be his main emphasis in the scheduled address. Therefore, the difference in initial attitudes between the two groups on the Viet Nam issue provides confirmation for the traditional selective-exposure hypothesis.

Our two extensions of that hypothesis, however, were not confirmed by the data. The failure to confirm our hypotheses regarding secondary attitudes and source credibility cannot be attributed to a restricted range of attitude on the concepts measured or to selection of inappropriate concepts for study. As noted in Table I, the range of attitudes on all of the concepts was comparatively large. The attitude change results indicate that at least six of the concepts chosen for measurement were relevant to the speech as delivered. The possibility exists, of course, that either our subjects were atypical or that this particular communicative event, because of its national importance, overcame our hypothesized selective-exposure effects. Both of these possibilities cause us to be hesitant to discount the hypotheses even though they failed to receive statistical support. In addition, the non-significant differences between the initial attitudes of the exposed and unexposed subjects on nine of eleven concepts were in the direction predicted by our second hypothesis. These three factors, as well as the theoretical justification of the hypotheses advanced previously, suggest that further research designed to test these hypotheses concerning selectivity in exposure to mass media communication would be appropriate.


1. Lipset, S. M., Lazarsfeld, P. F., Barton, A. H., and Linz, J. "The Psychology of Voting: An Analysis of Political Behavior," in G. Lindzey (ed.) Handbook of Social Psychology, Cambridge, Mass.: Addison-Wesley, 1954, Vol. 2.

2. Cf. Osgood, C. E.; Suci, G. J., and Tannenbaum, P. H. The Measurement of Meaning, Urbana, Ill.: University of Illinois Press, 1957; and Festinger, L. A Theory of Cognitive Dissonance, Stanford, Cal.: Stanford University Press, 1957.

3. See, for example, Brodbeck, M. "The Role of Small Groups in Mediating the Effects of Propaganda," Journal of Abnormal and Social Psychology, 52:166-170 (March, 1956), Cartwright, D. "Some Principles of Mass Persuasion," Human Relations, 2:253-267 (July, 1949); Ehrlich, D., Guttman, I.; Schonbach, P.; and Mills, J. "Postdecision Exposure to Relevant Information," Journal of Abnormal and Social Psychology, 54:98-102 (January, 1951); Mills, J.; Aronson, E. and Robinson, H. "Selectivity in Exposure to Information," Journal of Abnormal and Social Psychology, 58:250-253 (March, 1959); Rosen, S. "Postdecision Affinity for Incompatible Information," Journal of Abnormal and Social Psychology, 63:188-190 (January, 1961); Star, S., and Hughes, H. "Report on an Educational Campaign: the Cincinnati Plan for the United Nations," American Journal of Sociology, 55:389-400 (January, 1950); and Lazarsfeld, P.; Berelson, B.; and Gaudet, H. The People's Choice, New York: Duell, Sloan & Pearce, 1944.

4. Diab, L. N. "Studies in Social Attitudes: II. Selectivity in Mass Communication Media as a Function of Attitude-Medium Discrepancy," Journal of Social Psychology, 67:297-302 (December, 1965).

5. McGinnies, E., and Rosenbaum, L. L. "A Test of the Selective-Exposure Hypothesis in Persuasion," Journal of Psychology, 61:237-240 (1965).

6. Ibid., p. 240.

7. Anderson, K. and Clevenger, T., Jr. "A Summary of Experimental Research in Ethos," Speech Monographs, 30:59-78 (June, 1963).

8. McCroskey, J. C. "Experimental Studies of the Effects of Ethos and Evidence in Persuasive Communication," Unpublished Doctoral Dissertation, Pennsylvania State University, 1966; and McCroskey, J. C. "Scales for the Measurement of Ethos," Speech Monographs, 33:65-72 (March, 1966). The scales employed for the twelve substantive issues were harmful-beneficial, good-bad, wrong-right, fair-unfair, negative-positive, and wise-foolish. The scales employed for source credibility were (authoritativeness dimension) informed-uninformed, unqualified-qualified, reliable-unreliable, worthless-valuable, intelligent-unintelligent, and inexpert-expert; and (character dimension) sinful-virtuous, pleasant-unpleasant, dishonest-honest, nice-awful, friendly-unfriendly, and selfish-unselfish.

9. On the basis of the work of Boneau, it would appear that with the size of "n" included in this study violation of the assumption of normality would not have a major effect on the results.

Click Here To Go Back To PERIODICALS