The Relative Effectiveness of One-Sided and Two-Sided Communication For Mass Transit Advertising

ABSTRACT - This research investigates the relative impact of one-sided and two-sided communication on ridership intentions, attitudes toward transit features, and advertisement specific variables. The influence of varying numbers of claims is also studied. Two-sided communication was more effective than one-sided communication in increasing copy believability.

Citation:

Linda L. Golden and Mark I. Alpert (1978) ,"The Relative Effectiveness of One-Sided and Two-Sided Communication For Mass Transit Advertising", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 12-18.

Advances in Consumer Research Volume 5, 1978      Pages 12-18

THE RELATIVE EFFECTIVENESS OF ONE-SIDED AND TWO-SIDED COMMUNICATION FOR MASS TRANSIT ADVERTISING

Linda L. Golden, The University of Texas at Austin

Mark I. Alpert, The University of Texas at Austin

ABSTRACT -

This research investigates the relative impact of one-sided and two-sided communication on ridership intentions, attitudes toward transit features, and advertisement specific variables. The influence of varying numbers of claims is also studied. Two-sided communication was more effective than one-sided communication in increasing copy believability.

INTRODUCTION

For several years now the advertising prohibition against saying something negative about the advertised brand has not been strictly adhered to. Some advertisers have been "disclaiming" certain trivial characteristics of their products in conjunction with positive claims. Volkswagen, for example, used this technique to poke fun at itself while it was actually making positive claims with regard to the product.

In a disclaiming situation, an advertiser makes positive statements about characteristics that are determinants of product use, but does not claim that the product performs well on certain characteristics that are not determinants of product use. Previous research indicates that disclaiming may tend to increase the credibility of an advertisement (Settle and Golden, 1974). This may then result in a more effective advertisement.

Disclaiming in an advertisement may be viewed as providing the audience (consumers) with a two-sided argument with respect to the advertised product. For years communication research has investigated the attitudinal impact of two-sided arguments; however, the research area has not been developed in an advertising context. The focus on this research is to investigate the relative effectiveness of one and two-sided messages for mass transit advertising according to hypotheses derived from the communication research.

One-sided and Two-sided Communication Research

Communication researchers offer slightly different definitions of two-sided arguments (McGuire, 1954; Hovland, 1954; Hovland, Lumsdaine and Sheffield, 1949; Jones and Girard, 1967). Unlike other authors' definitions, Hovland's (1954) definition of two-sided arguments explicitly states that the communicator takes into account both sides of an issue, but he is himself in favor of one side. This has been described by Klapper (1949) as partial impartiality. Even though the advertiser may say something unfavorable about his brand or something favorable about the competitor, overall the advertiser will present his product as the one the consumer should buy. The advertiser is, indeed, in favor of one side. Thus, for purposes of this research Hovland's (1954) definitions are used. A one-sided argument is an argument confined to one side of an issue (Hovland, 1954).

The earliest studies in this area were designed primarily to investigate the effects of two-sided presentations. In these studies, the communicator impartially presented both sides of an argument without favoring either side. The general conclusion derived from these studies is that when one is successively exposed to first one side and then the other of a controversial subject, the typical result is that the individual is left at approximately his/her initial position (Sims, 1938; Schanck and Goodman, 1939).

The earliest experimentation explicitly directed to the investigation of the comparative effects of one-sided and two-sided communications was conducted during World War II by Hovland, Lumsdaine and Sheffield (1949). No main effect of direct attitude change was found in this study, but there were interactions with initial favor-ability such that one-sided communications were more effective for those initially in favor of the conclusion and two-sided communications were more effective for those initially opposed to the conclusion. There was also a significant interaction with education such that two-sided communications were more effective for high school graduates and one-sided communications were more effective with subjects who had not graduated from high school. Later studies by Janis, Lumsdaine and Gladstone (1961), Lumsdaine and Janis (1953), and Paulson (1954) also indicated that one-sided and two-sided arguments were about equally effective over-all in producing direct attitude change.

Thistlethwaite and Kamenetsky (1955) and Thistlethwaite, Kamenetsky and Schmidt (1956) investigated the attitudinal effects of refutation of opposing arguments rather than simple mention of opposing arguments. For some groups tested, there were no significant differences between the speeches with refutation and those without. For others, the refutation speeches had more influence. The authors concluded that the speeches with mention and refutation of opposing arguments had the effect of strengthening opposing attitudes. They suggest that listeners apparently discounted the speeches with refutation as "phony" attempts to seem impartial.

All of these studies seem to suggest that mention of opposing arguments should be handled with caution. The only groups that seem more positively affected by two-sided messages were those initially opposed to the conclusion and those of higher educational levels. Even these groups did not make large changes in attitudes. Two-sided messages, however, do have a specific place in the communicator's organizational framework. They can serve to "immunize" receivers against contradictory information in later situations (Lumsdaine and Janis, 1953; McGuire, 1961; McGuire, 1962; McGuire and Papageorgis, 1961).

Development of Hypotheses

Direct generalizations from communication research to advertising are restricted. The topics of the persuasive communications presented in the communication research were of a controversial nature. It is doubtful that the topic of many messages featured in an advertisement for consumer package goods could be considered controversial. However, advertisements for some non-traditional products or services such as birth control, welfare, and possible mass transit have topics which may be considered controversial. Further, the dependent variable in the communication literature is attitude change. The objective of advertising is to influence, in the long-run, not only attitudes but ultimately behavior. However, given the relatively high level of educational attainment of persons likely to switch from a car to the use of mass transit (Alpert and Davies, 1975) and the relative degree of controversy surrounding mass transit compared to consumer package goods, two-sided communication is a realistic promotional tool for mass transit to explore.

The purpose of this research is to investigate the relative impact of one and two-sided communications upon ridership intentions and attitudes toward mass transportation for individuals in a sample whose demographic characteristics approximate those of "potential switchers" to mass transit identified in previous research (Alpert and Davies, 1975). Because of the relatively high level of educational attainment by persons in this sample the following hypothesis was developed:

Two-sided messages will be more effective than one-sided messages in producing changes in attitudes toward mass transit and ridership intentions.

Since the number of positive claims included in the message may influence the effectiveness of the disclaimers, the effects of number of attributes in the copy were also tested. One hypothesis was generated to test this proposition:

The changes produced in ridership intentions and attitudes toward mass transit by one and two-sided messages will be influenced by the number of positive attributes included in the message.

METHODOLOGY

Presentation of both one and two-sided experimental manipulations requires selection of both determinant and non-determinant attributes for mass transportation. In the two-sided manipulations, the product does not claim to possess the non-determinant attributes, but does claim to possess the determinant attributes. For transportation, determinant attributes are those attributes of a product which determine the consumer's modal choice.

Previous research in the city studied had identified determinant attributes for potential switchers. The five most determinant attributes for which the bus was rated superior to a private car were selected for use in this research. These were: economy, freedom from parking problems, freedom from repairs, low energy use per passenger, and low pollution per passenger. Given a bus's perceived superiority on these features, it is likely that advertising which asserts these as advantages might be at least believable. The selection of the non-determinant attributes required additional testing, since it was necessary that the non-determinant attributes be believable both as positive claims (one-sided) and disclaimers (two-sided). The non-determinant attributes from previous research (e.g., "ability to read," "fun to drive") could not realistically be used for both positive and negative claims, because the image of one mode was perceptually superior.

The determinancy of fifteen potentially non-determinant attributes was tested on a sample of one-hundred university students who possessed characteristics closely approximating those of potential switchers. The results indicated that the attributes colorful interior and long windows would be suitable as non-determinant attributes for both the one-sided and two-sided manipulations. These attributes were rated as relatively unimportant transportation features for which cars and buses do not differ.

Attitudes toward both the bus and the car on these determinant and non-determinant variables (freedom from repairs, freedom from parking problems, low energy use per passenger, low pollution per passenger, economical, colorful interior, and long windows) constituted one set of dependent measures. Another set of dependent measures investigated respondents' affect toward driving a car and riding a bus. A third set of dependent measures investigated ridership intentions for specific trip purposes: to work or school and for shopping or personal business. The final set of dependent measures was designed to investigate attitudes toward the copy itself: likelihood of reading all the copy, believability, perceived quantity of information, usefulness of information, and general affect.

A pilot study was administered to a sample of 110 subjects whose characteristics closely approximated those of potential switchers, to test alternative ways of presenting the one-sided and two-sided communication formats and placement of dependent variables. A subject received one of several experimental manipulations followed by the dependent variables tentatively selected for use in the final instrument. The order of presentation was either: (1) experimental manipulation, dependent variables, media questions, or (2) experimental manipulation, media questions, dependent variables. The message formats (experimental manipulations) tested varied in their presentation of the attributes of the bus. The attributes were listed in a column below several sentences of copy, and the bus was described in one of three ways on each of the attributes. In one treatment, the bus was given a rating of either "superior" or "inferior" on the attributes. The one-sided treatment identified the bus' performance on all of the attributes as "superior." The two-sided treatment identified the bus performance on the determinant attributes as "superior" and as "inferior" on the non-determinant attributes. A second treatment followed the same general format, but replaced the adjective "superior" with "good" and "inferior" with "fair." The third treatment used check marks () beside the attributes under columns labeled either "bus gives you" or "bus doesn't give you." The one-sided treatment did not contain the column "bus doesn't give you" and checked each attribute under the column labeled "bus gives you." The two-sided treatment varied in that it checked non-determinant attributes under "bus doesn't give you." The results of the pilot indicated that the use of check marks provided a slightly stronger manipulation than any of the other two treatments tested. There were no significant differences for the alternative placements of the dependent measures.

The Final Instrument

The final instrument used an after-only design with control and contained five sections. The first section presented the respondent with one of ten different experimental manipulations. The respondent could receive either a one- or a two-sided communication containing either three, four, five, six, or seven attributes. The attributes were always presented in the same order, even though the number of attributes could vary. The non-determinant attributes were always the second and third attributes presented to the respondent. The experimental manipulation was printed on heavy glossy paper on a separate page in order to simulate an advertisement as closely as possible. A cover page told the respondent that the following page contained part of an advertisement and to please read it carefully. Figure 1 exhibits a two-sided experimental manipulation with seven attributes.

FIGURE 1

EXPERIMENTAL MANIPULATION FOR TWO-SIDED MESSAGE WITH SEVEN ATTRIBUTES

The second section of the instrument contained five questions concerning the subject's reactions to the copy. These questions were designed to ascertain the subject's likelihood of reading the copy in a magazine, the credibility of the copy, the information provided, the usefulness of the information, and the general attitude toward the copy. Responses were elicited according to a seven-point horizontal scale with one indicating the negative extreme.

The third section of the instrument obtained information regarding the subject's media habits. Information concerning the extent and nature of the subject's use of newspapers, radio, and television was elicited.

In the fourth section of the instrument, subjects were asked to indicate how likely they would be to purchase the product described in the experimental manipulation. In addition, information concerning the extent to which the subject felt the product possessed each of the seven attributes which could appear in the experimental manipulations was obtained. Subjects indicated their responses according to a seven-point horizontal scale with one representing "not at all" and seven representing "very much."

The final section of the instrument obtained demographic and personal information. Information regarding age, marital status, sex, employment status, household size, income, education, race, living situation, and number of automobiles owned was collected. On the last page of the instrument, the subject had the opportunity to request a summary of the survey results.

The final instrument was pre-tested for clarity of presentation on a sample of twenty subjects whose characteristics approximated those of potential switchers. Minor wording changes were made in the instrument as a result of the pre-test. The control group instrument was identical to that containing the experimental manipulation, except that the five questions directly regarding the advertisement (which was absent) were deleted.

Sample Selection and Administration

The criterion for the selection of subjects for the instrument was the possession of characteristics approximating those for potential switchers. In the survey area, potential switchers to mass transit had been found to be relatively younger, have smaller households, are more likely to be full-time or part-time students (although most are non-students), and they are more likely to shop and work in the downtown area than are those less likely to switch to mass transit (Alpert and Davies, 1975).

Distinct areas of a medium-sized southwestern city were identified which contained a relatively high proportion of individuals possessing the characteristics of potential switchers. An enumeration of households in these areas was obtained from Cole's Directory. In order to obtain a sample of 1,500 individuals, computer-generated random numbers were used to identify every nth person to be included in the sample frame. Only residents, not businesses, were counted when identifying potential subjects. Further, the sample was restricted to households within one-quarter mile of a current bus route, so that intention to ride the bus could be realistically measured.

Having identified the potential respondents, interviewers then began contacting by telephone. Interviewers were to ask specifically for the person whose name appeared on their calling list. Upon contact, the interviewer first gave his or her name and then requested their assistance in a consumer attitude survey being conducted by members of the Department of Marketing. When an individual agreed to participate in the study, he or she was told that they would receive the survey within a week. The respondent was instructed to please fill out the survey completely and return it at the earliest convenient time in the enclosed return envelope. Subjects were randomly assigned to treatments at the time of mailing. A letter of appreciation was included with the survey which contained the telephone number of the Department of Marketing so that the subject would have a contact point for any questions.

Statistical Analysis

The final sample consisted of 292 usable surveys returned from respondents. There were no less than 20 respondents in each cell with a maximum of 44 in the control groups.

Descriptive statistics (Veldman, 1967) confirmed that the demographic composition of the sample was highly similar to the characteristics of "potential switchers" to mass transportation derived from previous research. A second preliminary analysis performed on the data was a discriminant analysis (Veldman, 1967) to determine if respondents assigned to alternative treatments differed significantly on demographic dimensions. Three separate analyses were run: (1) comparison of respondents assigned to one-sided or two-sided treatments, (2) comparison of respondents assigned to three, four, five, six, or seven claims, and (3) comparison of respondents in each of the eleven treatment levels (including control groups). In each of these analyses, the ten demographic questions constituted the independent variables. There were no significant differences between respondents according to demographic variables for any of the above three analyses. Thus, respondents appear to have been randomly assigned to treatments on this dimension.

A final preliminary analysis was a descriptive analysis (Veldman, 1967) of the sample's ridership of the bus. Ninety-nine percent of the respondents used their car for trips to shopping or personal business. Sixty-four percent of the respondents used their car for trips to work or school; however, twenty-two percent of the respondents did not respond to this question since they did not work or go to school. Only four percent of the sample used the bus at all in the last four weeks. Thus, the sample is composed of individuals who use their car as their primary mode of transportation.

In order to compare the effectiveness of each of the experimental manipulations (advertisement treatments) against a control group, individual t-tests were performed on each of the twenty dependent variables for the respondents receiving a bus treatment. In addition, the data were submitted to two-way analysis of variance (Veldman, 1967) for the effects of communication type (one-sided versus two-sided) and the number of claims (three, four, five, six, seven). These results are reported in Tables 1 and 2 respectively, which appear in the next section. The data from the five advertisement specific dependent variables were submitted to two-way analysis of variance (Veldman, 1967) for each dependent variable separately in order to investigate the relative effects of communication type and number of claims. These results are presented in Table 3 of the next section.

RESULTS

Table 1 presents the results of the t-tests between the control means and the experimental manipulation means. The four most important dependent variables that can be used to evaluate the bus advertising treatments are the first four variables listed in Table 1. These measure the behavioral intentions toward the use of buses for trips to work or school (commuting) and for shopping or personal business, both over the short-run (How likely are you to ride the bus in the next month?) and "for most of your trips." As can be seen in Table 1, neither the one-sided nor the two-sided advertisement style was able to achieve any strong pattern of impact on people's behavioral intentions, which remained near the low end of the seven-point scale.

As can be seen from the cells in Table 1, there are 200 possible comparisons between the dependent measure ratings given by persons exposed to varying treatments and those in the control group. However, out of 200 possible comparisons one would expect 10 "significant" differences due to error or sampling fluctuations, using the .05 level for type-I error and two-tailed tests. There were 14 experimental rating means which were significantly different from the control mean, and of these 14, nine experimental treatment means did better than the control mean. This is hardly a strong overall pattern of change in attitudes or ridership intentions toward mass transit. Further, all the significant treatment-control comparisons for the one-sided messages were greater than the control, while only two of the seven significant comparisons for the two-sided were greater than control. However, five of the seven significant one-sided comparisons were for non-determinant attributes (not previously salient to buses).

Table 2 indicates the significant effects obtained in the two-way analysis of variance performed on the bus variables. For four of the twenty dependent variables, communication type was shown to have a significant main effect. In all of these, one-sided communication produced a more favorable evaluation of the bus than did two-sided communication. Two of the significant dependent variables, colorful interior and long windows, were disclaimed in the two-sided treatment, so it is logical that the one-sided treatment would have more impact. The theory had hypothesized that since these features were not determinant attributes of modal choice, it would be better to "give up" some perceptions in these attributes in return for higher evaluations in terms of the determinant attributes that would be claimed. Only one attribute mentioned in the copy, low pollution per passenger, showed a significant main effect of communication type. However, the results are counter-theoretical as the one-sided was more effective than the two-sided communication. The one-sided message was also more effective for liking the bus as an alternative to the car.

The results for colorful interior and long windows probably simply indicate that respondents believed the experimental manipulation, but beyond these two variables only two dependent variables showed a significant main effect of communication type. This is one more than would be expected by chance (alpha = .05), but the results indicate a stronger impact of the one-sided message. On the basis of these results, the first hypothesis which states that two-sided communication will be more effective than one-sided communication is rejected.

As further noted in Table 2, the impact of the number of claims produced significant between group variation for three of the twenty variables, which is more than would be expected by chance. A significant main effect for number of attributes was produced by: likely to ride the bus for shopping or personal business within the next month, enjoy driving car, and the perceived economy of the car. No distinct pattern emerges with regard to the optimum number of attributes. It is interesting to note that the number of attributes mentioned in the advertisement about the bus influenced the perceptions of the perceived economy of the car.

TABLE 1

SIGNIFICANT MEAN COMPARISONS FOR BUS AND CONTROL TREATMENTS

TABLE 2

ANALYSIS OF VARIANCE FOR BUS VARIABLES (SIGNIFICANT EFFECTS)

There was no significant interaction between number of claims and communication type, indicating that the number of claims had no significant impact on the dependent variables -- no matter which format (one-sided vs. two-sided) was used -- given the range of claims used (three to seven). Thus, the second hypothesis, which states that the number of claims will influence the effectiveness of the communication type used, was also rejected.

In order to obtain further information on the appropriateness of particular strategies, it may be useful to consider respondents' reactions to the advertisements themselves. It may be that one reason why attitudes toward the bus were not significantly influenced by the treatments is that the advertisements themselves were at fault. Table 3 depicts the results of the two-way analysis of variance for the ratings of advertisement variables.

There were five advertisement specific dependent variables: likelihood of reading all the copy if the advertisement appeared in a magazine, perceived believability, perceived quantity of information, perceived usefulness of information, and the extent to which the respondent liked the copy. There was a significant main effect of communication type for one of the dependent variables: believability. The results tend to indicate that two-sided communication was more believable than was one-sided communication.

There was a significant main effect for the number of attributes for three of the five dependent variables: believability, quantity of information and like copy. Six attributes produced the highest mean ratings for all three dependent variables. There was also a significant interaction effect for communication type and number of attributes for the dependent variable likeability of the copy. The highest mean rating appears for the two-sided communication with six attributes. For some numbers of claims, the one-sided communication appears to be more effective, and for other numbers of claims the two-sided communication was more effective.

CONCLUSIONS AND IMPLICATIONS

The experimental manipulation used in this study was a piece of advertising copy administered at one point in time via a static after-only with control experimental design. This one-time exposure to a portion of an advertisement may help to account for the lack of extensive significant impacts of ridership intentions or attitudes toward the bus. However, the manipulations were strong enough to produce significantly different effects for the advertisement specific variables. In the context of the limitations of the experimental design and the product under consideration, several conclusions can be drawn from the results of this study.

Two-sided communications were no more effective in producing changes in ridership intentions or attitudes toward attributes of the bus than were one-sided communications. In fact, there was not only a lack of overall pattern of positive attitude changes, but there appears to have been a greater proportion of negative effects on specific bus features advertised for two-sided communication. It may be reasonable to speculate that the effect of advertising public transportation in a two-sided communication format to those who have the option of private transit and feel generally negative toward buses is to evoke less positive evaluations of bus features than are normally the case. It is also possible that people may be reacting against a possible attempt to influence them to use this transportation mode by rating it less positively than when they are asked (without any exposure to advertising copy) to indicate a behavioral commitment to using it. In general, attitudes toward the bus may be so negative (even for potential switchers) that even the mildest of disclaimers only serves to reinforce opposing attitudes toward the bus.

TABLE 3

ANALYSIS OF VARIANCE FOR RATINGS OF ADVERTISEMENT VARIABLES (SIGNIFICANT EFFECTS)

Communication research has shown that two-sided communication is likely to be more effective for higher educated persons and those initially opposed to the conclusion. Although the sample was relatively higher educated and did not tend to ride the bus, neither one-nor two-sided communication produced significant attitude or ridership intention change. It may be argued that it is unfair to expect much change in overall attitude toward the bus, given only one exposure to a partial advertisement. This is particularly a problem given the major perceived disadvantages of buses in terms of convenience, flexibility, safety from dangerous people, and other determinant attributes found in prior research in this area, and not covered in this advertisement. It may be that the key to changes in rider-ship intentions and attitudes toward the bus lies in changes in the system and not in the nature of the promotional message used for this product. Attitudes may be so negative that a product change may be required to shift them, because the product may not offer significant advantages (possibly it offers significant disadvantages) over the most popular mode of transportation: the car.

Two-sided communication does appear to produce more believable copy than does one-sided communication. However, the believability is such that it does not make attitudes more positive and may in fact cause attitudes to move in a negative direction toward the attributes for which positive claims are made.

The number of attributes described in the copy tended to be more critical to the effectiveness of the advertisement for this research than did the communication format employed. For attitudinal variables toward the bus, there was no clear pattern of the optimum number of attributes. However, this research indicated that incorporating six attributes in the copy produced the highest ratings on advertisement related variables. Consumers tended to believe the advertisement more, perceived more information and liked the copy better when it contained six attributes. In addition, liking the copy may be a function of both the number of attributes and the communication type. Clearly, the operational relevance of this information depends upon the objectives to be achieved by the advertisement.

In order to more completely investigate the relative impact of one- and two-sided advertising, this research needs to be extended to more products, more independent and dependent variables, and to potentially stronger but realistic manipulations, over time. From the results of this research it appears that two-sided communication may be less effective than one-sided communication in producing positive attitude change toward product features. However, two-sided messages appear no more or less effective in influencing ridership intentions. Further testing is required with a myriad of situations and variables before any definitive relative effectiveness information can be discerned. What is likely to be most effective for mass transit is changes in the system along critical determinant attributes.

REFERENCES

M. Alpert and S. Davies, The Marketing of Public Transportation: Method and Application, Research Report 19, Council for Advanced Transportation Studies (The University of Texas at Austin, 1975).

L. Golden, "Attribution Theory Implications for Advertisement Claim Credibility," Journal of Marketing Research, 1(1977), 115-117.

C. Hovland, "Effects of the Mass Media of Communications" in G. Lindzey, ed. Handbook of Social Psychology, Vol. 2 (Cambridge: Addison-Wesley, 1954).

C. Hovland, A. Lumsdaine, and F. Sheffield, Experiments in Mass Communication: Studies in Social Psychology in World War II Vol. 3 (Princeton: Princeton University Press, 1949).

I. Janis, A. Lumsdaine and A. Gladstone, "Effects of Preparatory Communications on Reactions to Subsequent News Events," Public Opinion Quarterly, 2 (1951), 487-518.

E. Jones and H. Gerard, Foundations of Social Psychology (New York: John Wiley and Sons, 1967).

J. Klapper, The Effects of Mass Media (New York: Columbia University Bureau of Applied Social Research, 1949).

A Lumsdaine and I. Janis, "Resistance to 'Counterpropaganda' Produced by One-Sided and Two-Sided 'Propaganda' Presentations," Public Opinion Quarterly, 17 (1953), 311-318.

W. McGuire, "The Nature of Attitudes and Attitude Change," in G. Lindzey, ed. Handbook of Social Psychology Vol. 3 (Cambridge: Addison-Wesley, 1954), 136-314.

W. McGuire, "Persistence of the Resistance to Persuasion Induced by Various Types of Prior Belief Defenses," Journal of Abnormal Social Psychology, 64 (1962), 241-248.

W. McGuire and D. Papageorgis, "The Relative Efficacy of Various Types of Prior Belief-defense in Producing Immunity Against Persuasion," Journal of Abnormal Social Psychology, 62 (1961), 327-337.

S. Paulson, "The Effects of Prestige of Speaker and Acknowledgement of Opposing Arguments on Audience Retention and Shift of Opinion," Speech Monographs, 21 (1954), 267-271.

R. Schanck and C. Goodman, "Reactions to Propaganda on Both Sides of a Controversial Issue," Public Opinion Quarterly, 3 (1939), 107-112.

R. Settle and L. Golden, "Attribution Theory and Advertiser Credibility," Journal of Marketing Research, 1 (1974), 181-185.

V. Sims, "Factors Influencing Attitude Toward the TVA," Journal of Abnormal Social Psychology, 33 (1938), 34-36.

D. Thistlethwaite and J. Kamenetsky, "Attitude Change through Refutation and Elaboration of Audience Counter-arguments," Journal of Abnormal and Social Psychology, 51 (1955), 3-12.

D. Thistlethwaite, J. Kamenetsky and H. Schmidt, "Factors Influencing Attitude Change through Refutative Communication,'' Speech Monographs, 23 (1956), 14-25.

D. Veldman, Fortran Programming for the Behavioral Sciences (New York: Holt, Rinehart and Winston, 1967).

----------------------------------------