Selling a City: an Experimental Study of the Communication Effects of Message Tone

Richard F. Yalch, University of Washington (student), University of Washington
Margaret C. Dempsey,
ABSTRACT - Residents and nonresidents of Seattle were exposed to persuasive appeals varying in message tone but all urging them to live within the city limits. The results support the use of positively and neutrally worded messages, but suggest that negatively worded appeals be used only under limited circumstances.
[ to cite ]:
Richard F. Yalch and Margaret C. Dempsey (1978) ,"Selling a City: an Experimental Study of the Communication Effects of Message Tone", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 5-11.

Advances in Consumer Research Volume 5, 1978      Pages 5-11


Richard F. Yalch, University of Washington

Margaret C. Dempsey (student), University of Washington


Residents and nonresidents of Seattle were exposed to persuasive appeals varying in message tone but all urging them to live within the city limits. The results support the use of positively and neutrally worded messages, but suggest that negatively worded appeals be used only under limited circumstances.


Despite widespread praise because of its many favorable attributes, Seattle, as many other large cities, experienced a constant decline in its population and lost businesses to its suburbs during the past few years. Government officials attributed this partially to a negative image of urban living presented in news media accounts of urban problems and they sponsored a series of television commercials extolling the virtue of living within the city limits. The objective was to present "a number of really positive things going on around town the people never hear about" (Seattle Turns to Net TV, 1977). Concurrently, a county transportation agency tried to increase the demand for public transportation with an advertising campaign criticizing individuals who commute by automobile. The campaign was sufficiently negative toward automobile drivers (the slogan was "I'm no dummy, I take the bus") to motivate a public protest from the Executive Vice President of the Automobile Club of Washington which included his recommendation that the agency switch to a "positive" appeal (Dummies Who Drive, 1977). Although there may be a preference for a positive approach, the transportation agency's campaign and other critical messages seem justifiable if they are more effective in changing behavior than positive approaches.

The attitudinal and behavioral effects of these contrasting approaches to demand modification are usually unclear because of many environmental differences, and the controversy over the choice of wording seems best settled using controlled experimentation. Unfortunately, prior research on the relative effectiveness of positive and negative appeals has not yielded conclusive findings. [The wording or tone of a message is determined by the proportion of the message content that is positive (i.e., discusses the benefits of acquiring some good, service, or idea), compared to that which is negative (i.e., mentions the disadvantages of not acquiring the product or acquiring a substitute). Thus, a positive appeal might stress how commuting by bus is relaxing while a negative message would emphasize the tension caused by rush hour traffic jams. This definition is consistent with Lucas and Benson's (1929) view that "avoidance is the basis of negative appeal in advertising, (p. 275), but is somewhat different from that offered by Wheatley and Oshikawa (1970).] Although most studies favor the negative appeal (Evans, Rozelle, Lasater, Dembroski and Allen, 1970; McArdle, 1972; Powell and Miller, 1967; Wheatley and Oshikawa, 1970), several have found no differences (Leventhal and Perloe, 1970; Lucas and Benson, 1929) and one has found that positively worded appeals are more effective than negatively toned ones (Yalch and MacLachlan, in review).

Further, a review of this literature suggests that contributions to our understanding of the message tone effect can be made in at least four areas.

One problem is that many studies used a fear appeal as the negative message (cf., Evans et al., 1970; Powell and Miller, 1967; Wheatley and Oshikawa, 1970), and it is possible that its enhancement of persuasion was caused more by the greater anxiety experienced by the subjects during their exposure to this message than to the negative tone. The finding by Sigall and Helmreich (1969) that an anxiety-arousing situation enhances persuasion supports this suggestion. Thus, it would be desirable if all the messages in an experiment on message tone produced little anxiety or, at least, that the resultant anxiety levels were equivalent for all the messages.

A second limitation of the extant studies is that most used only two levels (one positive and one negative) of message tone (an exception is Powell and Miller, 1967). Many researchers have commented on the high probability that a message characteristic will have a nonlinear relationship with the audience's attitudinal response because of enhancing and attenuating effects on the intervening stages in the communications process. For example, Ray and Wilkie (1970) postulate that there is an inverted-u curve relationship between the level of anxiety aroused by a message and the amount of attitude change caused by it. Clearly, research designs incorporating more than two levels of message tone are needed to test for the possibility that neutrally worded messages are more effective than either positive or negative appeals.

A third concern is with the measures taken to assess the effects of exposure to messages varying in tone. In line with the previously stated possibility that the effects of tone have been confounded with anxiety arousal, it would seem useful to take a measure of the acute anxiety felt during the exposure to the message. Also, the effectiveness of the tonal manipulation should be demonstrated and carryover effects to perceptions of other message and source characteristics should be checked. Further, in consideration of wide support for McGuire's (1976) multimediation model of the communication process, the audience's comprehension of the arguments used in the message as well as the attitudinal effects should be assessed. Assessment of these mediating factors is critical in formulating explanations for why one message is more persuasive than another and in developing recommendations for actual message construction.

The fourth area requiring additional research is to evaluate the moderating effects of individual personality factors on the effects of positive and negative appeals. Wheatley and Oshikawa (1970) found that a positive appeal was more effective when the audience had a high chronic anxiety level, but that a negative, anxiety-arousing appeal was more persuasive with a low chronic anxiety audience. Subsequent reanalysis of their data, however, revealed that product ownership was a more critical factor than initial anxiety in determining the effectiveness of the negative message (Wheatley, 1971). It would seem worthwhile to explore this point further using a different experimental setting as the theoretical and empirical support for the moderating effect of chronic anxiety is quite substantial (see Wheatley and Oshikawa, 1970).

These four research questions were explored in an experimental study of the effect of message tone on an audience's comprehension and acceptance of a series of arguments favoring one's living within the city limits of Seattle. The confounding effect of anxiety-arousing cues within a negative appeal was minimized by the topic selected and pilot testing of the messages. Nevertheless, a postexposure measure of felt anxiety was administered to verify that this was not an important factor. A neutral message (neither positive nor negative) was included in the design to test for curvilinear relationships, and aided recall as well as attitude questions were asked to evaluate the messages' effects. Finally, in consideration of prior theorizing about the importance of individual characteristics as moderating variables, the participants' chronic anxiety and product ownership (as measured by place of residence) were included among the personal characteristic questions.



Participants in the experiment were 184 adults contacted in public places such as bowling alleys, apartment buildings, government offices, and hospitals. A non-student sample was chosen in a nonlaboratory setting in order to enhance the external validity of the results. An effort was made to recruit an equal number of persons residing in the city and its suburbs to test whether the effects of the appeals varied with the initial opinion of the audience.


Each participant was approached and asked if he or she would cooperate in a university research project. Those agreeing were given a booklet containing a questionnaire made up of several parts and one of the four written messages. Participants were randomly assigned to treatments by sorting the booklets using a random number table before the interviewers went into the field.

The initial part of the booklet consisted of a chronic anxiety scale to assess the participants' general anxiety level. This was followed by a description of the source of the message and the message itself, all on separate pages. After reading the message, the participants answered six pages of questions. These measured their feelings while reading the message (acute anxiety), their evaluations of the message and source (manipulation checks), their understanding of what the message said (reception), their agreement with a set of statements about living in Seattle and its suburbs (post-message attitudes), and their personal characteristics. Upon completion of the booklet, the participants were thanked for their assistance and debriefed on the purpose of the study and the need to expose them to a specially prepared message.

Independent Variable

The effects of message tone were studied using three experimental (positive-neutral-negative) and one control groups, each consisting of 46 participants. Prior to the final study, a sample of city and suburban residents ranked a list of reasons for selecting where to live (e.g., quality of schools and nearness to work). The reasons ranked as most important were converted into a set of statements about the advantages of living in the city and the disadvantages of living in the suburbs. Message tone was controlled by varying the proportion of city advantages and suburban disadvantages included in the appeal.

Positive appeal. The positive appeal argued entirely for the advantages of living in Seattle. The following is an excerpt from it:

When you buy a home in Seattle you get a lot of advantages...sewers, streets, utilities, and good schools are already available. Taxes in Seattle are lower than in the suburbs, which are still in the process of establishing these vital services.

The message's other arguments included the variety of housing available in the city, better public transportation, and time saved commuting to work. It began and concluded by praising those persons still living in the city.

Negative appeal. The negative appeal argued primarily against living in the suburbs in contrast to the positive appeal's arguing for living in the city. The following excerpt was constructed to correspond to that above for the positive message:

When you buy a lot or house in the suburbs, you have to help pay for the construction of sewers, utilities, and schools...things long established in Seattle and paid for years ago. In other words, taxes are higher in the suburbs than in the city.

In addition, the message began and concluded by criticizing those individuals who had recently moved from the city to the suburbs.

Neutral appeal. A third message was developed to be neutral in tone to test for possible curvilinear effects. It was composed of statements that were less negative and less positive than those used in the other appeals. The following excerpt corresponds to those previously given:

There are both advantages and disadvantages to living in the city rather than the suburbs. Taxes are lower in the city than the suburbs because vital services like sewers, streets, utilities, and schools are already established.

In this case, the message was moderated by deleting the references to buying a home and the necessity of establishing services in the suburbs. Further, there was no mention of approval for those remaining in the city or disapproval for those moving to the suburbs.

Control appeal. Because of the after-only design of the experiment, a nonpersuasive message was developed for presentation to a no-treatment control group in order to assess the absolute persuasiveness of the three experimental messages. It contained none of the arguments used in the other messages. The following illustrates the content:

Business in Seattle has recently been clicking along at a fast pace. The products turned out in Seattle are diverse...jet aircraft, space hardware, medicine, and more recently, professional sports team. The new King County Domed Stadium is the home of the Seattle Seahawk professional football team and will soon become the home of the Seattle Mariners professional baseball team. The "Dome" has brought great excitement to the residents of Western Washington.

Although the topics discussed were related to Seattle, they were more representative of the metropolitan area than the city itself and were neutral with regard to whether it was better to live within the city limits or the suburbs.

The positive and negative messages were pilot tested on a sample of city and suburban residents to verify that they were perceived as significantly different in message tone but not on any other major message characteristic. The neutral and control messages were developed after the pilot test.


Chronic anxiety. A measure of the subjects' chronic or general anxiety was taken before they were exposed to the message. Although chronic anxiety should be insensitive to situational factors, it is desirable to assess it before exposure to potentially anxiety arousing information. Initial pilot testing revealed that the 47 question, Sarason Lack of Protection scale (Adams and Sarason, 1963, p. 245) used by Wheatley and Oshikawa (1970) was overly long for a field study. Therefore, the Taylor Manifest Anxiety scale (Taylor, 1953) was used in the present study. It consists of 20 statements such as "I feel anxious about something almost all the time" and "I don't like to face a difficulty or make an important decision," each rated from definitely agree to definitely disagree using a five-point scale. The participants' responses were distributed uniformly throughout the scale with the actual range being from 24 to 99 compared to the scale's total range of 20 to 100. The reliability for the entire scale was excellent (Cronbach's alpha = .92).

Acute anxiety. After responding to the chronic anxiety test and reading one of the messages, the participants reported their acute or temporary anxiety while reading the message. This section consisted of five semantic differential items (anxious-not anxious, confident-not confident, calm-tense, threatened-unthreatened, and nervous-relaxed). The responses were summed to form the acute anxiety scale (Cronbach's alpha = .89), and used to determine if the effort to equalize the anxiety-arousing aspects of the message was successful.

Message and source evaluations. In order to verify that the message tone manipulation had been successful and to determine if tonal changes would alter the audience's perceptions of the attributed source of the message, the participants were asked to respond to two sets of semantic differential items. The first set consisted of ten bipolar adjectives describing a message. These included three on message tone (positive-negative, optimistic-pessimistic, and threatening-not threatening), two on message clarity (confusing-understandable and clear-unclear), and five on other message characteristics (strong-weak, irrelevant-relevant, exaggerated-under-stated, interesting-uninteresting, and constructive-destructive).

The attributed source of the message (Mr. Jim Wilson, a recognized authority in the area of urban planning and development) was evaluated in the second set of adjectives on three credibility dimensions. These were expertise (knowledgeable-unknowledgeable, competent-incompetent, experienced-inexperienced, and expert-inexpert), trustworthiness (trustworthy-untrustworthy, subjective-objective, unjust-just, and biased-unbiased), and attractiveness (not likeable-likeable, pleasant-unpleasant, similar to you-dissimilar to you, and offensive-not offensive). The three summed responses were used to determine if differences in response to variations in message tone might be attributed to changes in source credibility, and, if so, which particular attributes were affected.

Reception. It has been postulated that differences in the effectiveness of persuasive appeals might be attributed to the mediating stages in the communications process. The possibility that messages differing in tone might vary in attention maintenance or ease of comprehension was tested by having participants report the message's position on a series of statements. For example, participants were asked whether the statement, "Taxes are lower in the suburbs than in the city," was definitely true, partly true, undecided, partly false, or definitely false according to what they thought the message said. As the correct response was always definitely true or definitely false, it was possible to score the participants by how well their recollection matched this and to sum these into an overall measure of their understanding of the message (Cronbach's alpha = .61).

Postexposure attitudes. The participants were asked to report their level of agreement (five-point scale labeled strongly agree and strongly disagree at the ends), with ten statements about the advantages and disadvantages of living in the city or the suburbs. For example, "I would recommend living in the city to a newcomer to the Northwest," was one of the statements. Several statements were not relevant to the message and served as filler items. Therefore, only six of the ten statements were summed to form the overall attitude toward living in the city (Cronbach's alpha = .82).

Personal characteristics. The final section of the questionnaire consisted of five demographic questions about the participant's age group, education level attained, place of residence, male or female, and total annual family income. These responses were used to test the success of the randomization procedure and to determine whether product ownership as determined by place of residence moderated the effects of the message tone manipulation.


The results of the experimental manipulation of message tone were analyzed in terms of the effects of exposure to the message on the participants' acute anxiety, reception of the arguments, and attitudes toward living in the city.


The randomization of the participants to treatment groups was tested using the responses to the chronic anxiety scale which was administered before any were exposed to a message and to the personal characteristics questions. An analysis of variance of the anxiety scores and chi square analysis of the personal characteristics across treatment and control groups revealed that the groups were mostly homogeneous (p > .10). The only variation from homogeneity was for education. As the slightly lower educational level of the neutral and control groups relative to the positive and negative message groups was not thought to be theoretically important, the assumption that the groups were randomly selected was accepted as valid.

Acute Anxiety

The possibility that the messages differed in anxiety arousal as well as tone was tested by comparing the participants' reported feelings during their exposure to it. Unexpectedly, there was a significant difference (F = 9.4, df = 2,122, p < .001), with the neutral message being significantly less anxiety arousing (M = 6.4, n = 44) than the positive (M = 9.0, n = 40) and negative message (M = 9.8, n = 40) according to a studentized Newman-

Keuls posterior comparison of the individual cells. However, because the negative message was not perceived as more anxiety arousing than the positive message, it was felt that comparisons of the postexposure differences in these groups would be a valid test of the effect of message tone independent of anxiety arousal differences.

Message and Source Evaluation

The participants' perceptions of the message are reported in Table 1. Substantial differences were found for the tonal attributions and these were all in the predicted direction. The positive and negative message did not differ on clarity or any other characteristic. However, the neutral appeal was thought to be significantly less clear, less understandable, and less interesting than the positive and negative appeals. This may reflect the participants' difficulty in reading a message that frequently changed message tone. Because of this difference, one must be cautious in interpreting the effectiveness of the neutral message used in this study.



The source evaluations are reported in Table 2. There were no significant differences in the expertise dimension, but the source associated with the negative appeal was viewed as less trustworthy and attractive than those giving the positive and neutral messages. The neutral message source was evaluated most favorably on all three dimensions, though significantly so only in attractiveness.



Reception and Attitude Measures

The participants' immediate recall of the message and postexposure attitudes toward living in the city are reported in the last columns of Table 3 and 4, respectively. The negative message was significantly less well recalled and also resulted in the least favorable attitude toward living in the city. In fact, in comparison with the control group, the negative appeal group gave evidence of a "boomerang" effect. That is, they shifted their attitudes in the direction opposite from the argued in the message. However, a t test of the attitudinal differences between the experimental and control groups revealed that none of the differences was significant.

Moderating Influences of Chronic Anxiety and Residence

Two audience characteristics, chronic anxiety and current place of residence, were included in the analyses of variance of the dependent measures to assess their affect on the tone-persuasion relationship. The chronic anxiety scale responses were used to divide the sample into three groups (high-moderate-low) and reported residency responses were used to determine who was or was not a resident of the city (resident-nonresident). Tables 3 and 4 report the cell means for this breakdown and Table 5 has the analyses of variance with the three messages, three levels of anxiety, and two levels of residency as the independent factors.

After considering the important influence of the participants' residency and initial anxiety, message tone had a marginally significant effect on reception and an insignificant effect on attitudes. Comparing the interaction terms showed that the moderating effect of anxiety substantially exceeded that of residency for the reception measure and slightly exceeded it for attitudes.

The prediction that a positive message would be more effective than a negative message for individuals with high anxiety was in the right direction for both reception (M = 33.0 vs. 27.4) and attitudes (M = 19.7 vs. 18.1) measures, but was statistically significant only for the reception measure (t = 3.53, df = 112, p < .001). [The statistical test used is that for planned comparisons between cells (Tables 3 and 4) with the error term being estimated from the analyses of variance (Table 5). See Winer (1971, pp. 384-388).] The other prediction, that a negative message would be more effective than a positive message for individuals with low anxiety, was also in the predicted direction for reception (M = 33.5 vs. 28.1) and attitudes (M = 19.1 vs. 18.1), but again the difference was only significant for the reception measure (t = 2.75, df = 112, p < .01).








The results of this study offer further evidence that in the absence of differences in anxiety-arousing content, negatively-worded appeals are not more effective than positive messages. In fact, the data favor the use of positive appeals over negative appeals. Four issues identified in reviewing prior research (confounding of anxiety arousal with negative wording, necessity of having three levels of message tone, using multiple measures of exposure effects, and reexamining the moderator role of audience characteristics such as chronic anxiety and product ownership), were investigated by exposing participants to persuasive appeals varying in message tone but all urging individuals to live within the city limits of Seattle.

The participants' message evaluations and reported feelings during their exposure demonstrated that the positive and negative messages were successfully constructed to differ in tone without differing in anxiety arousal. However, both were significantly more anxiety arousing than the neutral message, apparently because they contained some social approval and disapproval statements. Since the neutral message was also perceived as being somewhat more confusing than the other two message, it is not clear that all postmessage differences should be attributed to the tonal differences in the messages and therefore conclusions about curvilinear effects are tenuous. The only evidence supporting the use of neutrally-worded messages was a generally more favorable evaluation of the message source compared to the positive and negative source evaluations.

One of the two major differences between the positive and negative messages was a loss of credibility for the source giving the negative appeal. The loss of trustworthiness and attractiveness suggests that an audience more easily views a negative appeal than a positive message as a persuasive attempt. This loss of credibility could be a serious concern for advertisers though there may be ways to counteract it, such as by employing highly credible sources to deliver the message. For example, the practice of having government health and safety warnings include a reference to the agency responsible would seem to be a wise strategy for enlisting public cooperation.

The other important difference was that the audience exposed to the positive message comprehended and recalled the content better than those exposed to the negative message. The fact that this led to only slight differences in postmessage attitudes probably reflects the single exposure opportunity, the message topic, and the audience. As most individuals had committed themselves to living in or out of the city, it would seem unlikely that they would easily change their opinion. With a less firmly established attitude or with repeated exposures it is possible that a significant attitude change would have been observed.

The improved reception combined with the higher source credibility evaluation offer strong support for employing only positive and neutral appeals and avoiding negative messages. Two factors argue against rigorous adoption of this recommendation, however. First, the literature review revealed that negative messages are effective when combined with other message factors, particularly anxiety-arousing information. Similarly, if very strong negative arguments are available, it is possible that their ability to impede counter-argumentation would more than compensate for the undesirable effects of negative message tone. Under these circumstances, an advertiser would benefit by sponsoring the negative campaign.

Secondly, the effects of message tone were not uniform across audiences. The positive message was most successful with audiences initially opposed to the position being argued (nonresidents in this study), and with high anxiety individuals. Both of these groups would probably feel most threatened by the negative appeal which would result perhaps in avoidance (evidenced by the low message reception), source derogation (indicated by the unfavorable source evaluations), and counterargumentation. However, for those initially agreeing with the message's position (city residents), and low anxiety individuals, the negative appeal was as effective as the positive appeal and only slightly less effective than the neutral message. To the extent that an audience possesses any of these characteristics, an advertiser would not suffer by using a negativistic approach.

It should be noted that a reexamination of the relative importance of chronic anxiety and product ownership did not support a previous finding. Contrary to Wheatley's (1971) conclusion, a measure of chronic anxiety was a more important moderator variable than was a measure of product ownership (place of residence). There are several methodological differences between the two studies that could account for this. The present study used low anxiety-arousing appeals (all three messages yielded reported anxiety levels of less than 10 on a scale with a range from 5 to 25). The negative appeal in the original Wheatley and Oshikawa (1970) experiment, which served as the data source for Wheatley's (1971) conclusion, was designed specifically to be anxiety arousing. If all the participants are experiencing high acute anxiety because of their exposure to a threatening message than some of the communication effect attributable to individual differences in chronic anxiety could be obscured. Also, this study employed a different anxiety scale and used nonstudents in a field setting as participants which appeared to have increased the range of responses to the anxiety questions compared to the student subjects tested in a classroom in the previous study. An increase in the range of scores of an independent factor would tend to increase the likelihood that it would account for more of the variance in a dependent variable. Further, Wheatley's findings were based on attitude change scores rather than postmessage attitudes. It is possible that ownership is related to the reliability of attitudinal responses which would confound any attempt to assess the effect of the persuasive appeals. Additional research is needed to determine which of these differences account for the inconsistency in the conclusions of the two studies.


Elsie B. Adams and Irwin Sarason, "Relation Between Anxiety in Children and Their Parents," Child Development, 34 (March 1963), 244-5.

"Dummies Who Drive," Seattle Times, March 28, 1977, p. A13.

Richard Evans, Richard Rozelle, Thomas Lasater, Theodore Dembroski and Bem Allen, "Fear Arousal, Persuasion, and Actual Versus Implied Behavioral Change," Journal of Personality and Social Psychology, 16 (October 1970), 220-7.

Howard Leventhal and Sidney Perloe, "A Relationship Between Self-Esteem and Persuasibility," Journal of Abnormal and Social Psychology, 64 (May 1962), 385-8.

D. Lucas and C. Benson, "The Relative Values of Positive and Negative Advertising Appeals as Measured by Coupons Returned," Journal of Applied Psychology, 13 (June 1929), 274-300.

J. McArdle, "Positive and Negative Communications and Subsequent Attitude and Behavior Change in Alcoholics," unpublished doctoral dissertation, University of Illinois, 1972.

William McGuire, "Some Internal Psychological Factors Influencing Consumer Choice," Journal of Consumer Research, 2 (March 1976), 302-19.

Frederic Powell and Gerald Miller, "Social Approval and Disapproval Cues in Anxiety-Arousing Communications," Speech Monographs, 34 (June 1967), 152-9.

Michael Ray and William Wilkie, "Fear: The Potential of an Appeal Neglected by Marketing," Journal of Marketing, 34 (January 1970), 54-62.

"Seattle Turns to Net TV to Promote City Living," Advertising Age, February 21, 1977, p. 20 BW.

H. Sigall and R. Helmreich, "Opinion Change as a Function of Stress and Communicator Credibility," Journal of Experimental Social Psychology, 5 (January 1969), 70-8.

Janet Taylor, "A Personality Scale of Manifest Anxiety," Journal of Abnormal and Social Psychology, 48 (April 1953), 285-90.

John Wheatley, "Marketing and the Use of Fear- or Anxiety-Arousing Appeals," Journal of Marketing, 35 (April 1971), 62-4.

John Wheatley and Sadaomi Oshikawa, "The Relationship Between Anxiety and Positive and Negative Advertising Appeals," Journal of Marketing Research, 7 (February 1970), 85-9.

B. Winer, Statistical Principles in Experimental Design. New York: McGraw-Hill, 1971.

Richard Yalch and Douglas MacLachlan, "Persuasive Effects of Varying an Appeal's Tone, Supporting Evidence, and Source Credibility," in review.