Corporate rumor activity, belief and accuracy

OF RO Public Relations Review 150 (2002) 1–18 Corporate rumor activity, belief and accuracy 4 Nicholas DiFonzo a,∗ , Prashant Bordia b a 5 6 7 Department of Psychology, Rochester Institute of Technology, 18 Lomb Memorial Drive, Rochester, NY 14623, USA b University of Queensland, Brisbane, Australia Received 1 September 2001; received in revised form 1 October 2001; accepted 1 December 2001 8 9 DP 3 Abstract 19 1. Introduction 20 Rumors that are considered harmful or potentially harmful reach the ears of Public Relations (PR) officers and communications consultants almost once per week on average.1 Well-known examples of such rumors include the false report that Bubble Yum bubble gum contained poisonous spider eggs and the false allegations that Pop Rocks candy, when eaten with soda, would explode in the stomach.2 Stories that Bank of Boston was linked with organized crime presented intense PR challenges.3 False rumors that Prodigy taps private information from subscriber hard drives were widely discussed on the Internet.4 Less publicized but common internal corporate rumors include those connected with organizational change, such as layoffs, reorganizations, mergers, and changes in management.5 13 14 15 16 17 21 22 23 24 25 26 27 28 ∗ RE C 12 CO R 11 TE 18 An integrated model relating workplace rumor activity, belief, and accuracy is proposed and tested. Senior VPs of Communications from a sample of Fortune-500 corporations and CEOs of established public relations firms were surveyed regarding rumor episodes that they had experienced. Results confirmed previous research on the role of uncertainty, anxiety, and belief in rumor activity. In addition, a reduced sense of control mediated the effects of uncertainty on anxiety, and anxiety mediated the effects of importance on rumor activity. Evidence was found for the roles of group bias in how strongly a rumor is believed. Rumor activity was also implicated in the formation of more accurate rumors. The significance of these results for rumor theory and for Public Relations practitioners is presented. © 2002 Elsevier Science Inc. All rights reserved. 10 1 2 UN Corresponding author. Tel.: +1-716-475-2907; fax: +1-716-334-4529. E-mail addresses: nxdgss@rit.edu (N. DiFonzo), prashant@psy.uq.edu.au (P. Bordia). 0363-8111/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved. PII: S 0 3 6 3 - 8 1 1 1 ( 0 2 ) 0 0 1 0 7 - 8 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 OF 2 46 1.1. Rumor defined 37 38 39 40 41 42 43 44 47 48 49 DP 36 TE 35 Rumors are bits of information that are of interest to people, that are not accompanied by secure standards of evidence, and that are proposed for possible belief.12 The key characteristic of a rumor is doubt about its veracity. It is helpful here to contrast rumors with news; news RE C 34 CO R 33 UN 31 32 RO 45 Rumors such as these have many deleterious consequences but most often include damage to employee morale and trust, and sullied organizational reputations.6 The effects may be severe.7 For example, the false rumor that Tropical Fantasy fruit punch contained a substance that caused African American men to become sterile resulted in a 70% loss of sales.8 False rumors that Continental Illinois was filing for bankruptcy hastened its demise.9 Rumors (which turned out to be true) of a massive facility relocation caused prolonged and substantial work slowdowns.10 Understanding the processes involved in rumor episodes and managing rumors to minimize potential harm are therefore key knowledge and skill sets for PR practitioners.11 Three characteristics of a rumor episode that are of interest to PR practitioners, rumor participants, and communications researchers are how active the rumor is, how strongly people believe the rumor, and how accurate (versus distorted) the rumor has become. Although some empirical work has investigated each of these areas separately, none has explored how these characteristics are related. The present research investigated these three aspects of rumor in an integrated model. A sample of top-level corporate PR officers and PR consulting firm CEOs were surveyed regarding actual rumor episodes they had dealt with. These data were used to test and refine the model. We first define rumor and review theory and research informing the model. The hypothesized model is presented in Fig. 1. 29 30 Fig. 1. Path specification for the hypothesized model. OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 3 58 1.2. Rumor activity 59 We define rumor activity as referring to the frequency with which the rumor is transmitted and discussed. Previous research has investigated a similar construct, rumor transmission, defined as whether or not the rumor is transmitted.15 Obviously, rumors that are active are likely to be transmitted (and vice versa); therefore we will use the terms interchangeably in this discussion. Social scientific research on the generation and spread of rumors has identified four important predictors of rumor activity: uncertainty, anxiety, importance, and belief. Uncertainty refers to the psychological state of doubt about what current events mean or what future events are likely to occur.16 Uncertainty generates rumor activity.17 When information is not available from formal channels—such as the news media or official sources—uncertainty abounds and people compensate by generating, refining, and transmitting rumors.18 Rosnow’s meta-analysis of factors associated with rumor transmission showed a small to moderate (r = .19) average linear effect for uncertainty across five studies.19 As Heath put it: 56 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 DP 55 TE 54 RE C 53 If people are uncertain about their circumstances, they are likely to employ a variety of tactics to reduce that uncertainty. They prefer to seek information through established channels and from appropriate sources. If that effort is too costly in proportion to its rewards or is perceived unlikely to reduce uncertainty, rumor and a contrived sense of the company will suffice for members to believe they know what is going on.20 Anxiety (in the context of rumor) has been defined as “an affective state—acute or chronic— that is produced by, or associated with, apprehension about an impending, potentially disappointing outcome.”21 In other words, anxiety is dread and worry over potential negative events. Anxiety generates rumor activity. Transit strike and campus murder rumors that made respondents feel anxious were transmitted more often than rumors that did not.22 Subjects in a high anxiety manipulation condition in an experimental study needed less prodding to repeat a planted rumor.23 Finally, Rosnow’s meta-analysis of seven studies found a large average linear effect (r = .48) for the relationship between anxiety and rumor transmission.24 Importance refers to how significant the issue is to which the rumor pertains. Allport and Postman posited that importance was a necessary condition to rumor transmission.25 They noted, “an American citizen is not likely to spread rumors concerning the market price for camels in Afghanistan because the subject has no importance for him, ambiguous [uncertain] though it certainly is.”26 However, empirical evidence supporting importance as a necessary condition for rumor activity is mixed.27 Though two rumors of importance to subjects—planted in an organization by experimenters—spread easily, a third one was reportedly not spread CO R 52 UN 51 RO 57 has secure standards of authentication, rumors do not.13 That is, rumors are just like news, only without the confidence that accompanies an article in a respectable newspaper. It is also helpful to contrast rumors here with gossip. Rumors are usually about important or significant topics; gossip is about is usually about private affairs and is used to entertain or convey social mores.14 We distinguish rumors from legends and folklore also. Rumors are set forth to be believed (I heard that this division is being sold) while whether or not a legend is true (George Washington chopped down a cherry tree) is not as important as the underlying message (tell the truth). 50 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 OF 4 108 Hypothesis 1. Anxiety, uncertainty, importance and belief predict rumor activity. 109 121 Though these four variables have each been implicated in rumor transmission, their conceptual inter-relationship bears further elaboration. Consider importance. In an attempt to reconcile the conflicting results of the effect of importance on rumor activity, Rosnow reconceptualized importance as outcome-relevant involvement.36 This term refers to how closely a rumor is related to outcomes that are relevant to us and how much it evokes our caring/involvement. We note that the idea of caring/involvement is closely related—though not synonymous—with the emotional tension/dread that defines anxiety. Thus, importance ought to predict anxiety. Further, as already discussed, anxiety leads to rumor activity. Putting these two ideas together, we hypothesized that anxiety would mediate the effect of importance on rumor activity. That is, the issue is important to me, it evokes caring/involvement, and some degree of anxiety, so I participate in the rumor mill to relieve anxiety. This mediation route more clearly reflects Rosnow’s reconceptualization of importance. Conversely, to the extent that a rumor does not evoke caring/involvement/anxiety, it will be less likely to be transmitted. 122 Hypothesis 2. Anxiety mediates the effect of importance on rumor activity. 123 130 Next consider uncertainty. Uncertainty has been correlated with psychological discomfort and stress,37 which is conceptually adjacent to anxiety. Further, the relationship between uncertainty and psychological discomfort/anxiety appears to be mediated by a reduced sense of control.38 That is, I have questions about what current events mean or portend, this reduces my sense of control over events that affect me, and I therefore feel anxiety. Indeed, Bordia et al. found that a reduced sense of control mediated the relationship between uncertainty and psychological strain.39 Similarly, we posit that a reduced sense of control mediates the relationship between uncertainty and anxiety. 131 Hypothesis 3. Reduced sense of control mediates the effect of uncertainty on anxiety. 98 99 100 101 102 103 104 105 106 110 111 112 113 114 115 116 117 118 119 120 124 125 126 127 128 129 DP 97 TE 96 RE C 95 CO R 94 UN 92 93 RO 107 because of its importance.28 Similarly, students who rated the rumor topic as low in importance transmitted the rumor—that a student had been caught smoking marijuana during a final exam— more frequently.29 Nevertheless, importance in rumors among patrons of a city transit system— closed due to labor strike—was positively correlated with rumor transmission.30 And in recent field interviews investigating organizational rumors each rumor was associated with one or more areas of concern (i.e., of importance) to the collective (e.g., job security, job quality, pecking order).31 For example, rumors of an impending manufacturing plant shutdown—an event that would obviously affect job security—were widely disseminated and discussed. Belief refers to how much confidence is placed in the veracity of the rumor. Rosnow found a moderate average effect (r = .30) across six studies that correlated belief in rumor with transmission.32 During a strike by university faculty, rumors that were firmly believed were transmitted more than those in which there was less confidence.33 Similar results were reported in a field study during the aftermath of a student’s murder on the campus of the University of Pennsylvania.34 Rosnow et al. have posited that if people lack confidence in a rumor, they usually desist from passing it on, lest they lose credibility in the eyes of recipients.35 Someone who raises false hopes or causes unjustified fear can reliably expect social castigation. Putting these findings together gives us our first hypothesis regarding activity: 91 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 5 135 Hypothesis 4. Reduced sense of control and importance will predict anxiety. 136 1.3. Belief in rumor 137 A substantial amount of descriptive and correlational research points toward belief in rumors that are consistent with the hearer’s currently held biases, attitudes, and prejudices. This can be explained by the well-known tendency for people to selectively interpret new evidence so as to conform to already-held ideas.40 Racial stereotypes and animosities lay behind the prevalence of racist rumors.41 Belief in rumors of “waste and special privilege” (e.g., that government officials had free access to gasoline) was correlated with unfavorable attitudes toward the (WW II) wartime rationing program.42 If a rumor supports or accords with that which the actor already hold to be true—one’s biases, attitudes, and prejudices—the actor will assign greater credence to it. 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 DP 140 141 TE 139 Hypothesis 5. Agreement of the rumor with group biases, attitudes, and prejudices will predict strength of belief in the rumor. Rumor content and anxiety have long been associated. More anxious groups tend to produce rumors that accord with group biases, attitudes, and prejudices.43 For example, “Eleanor Club” rumors (e.g., that African American maidservants were found preening themselves using the “lady of the houses” combs, brushes and vanity) arose most during strenuous times and were invariably expressions of prevalent White biases, attitudes, and prejudices.44 Anxiety-producing rumors of young boys being castrated in shopping mall lavatories varied according to the race of the group within which the rumor circulated: in Black communities, the perpetrator was rumored to be White, while in White communities, he was Black.45 Anxiety should therefore lead to greater agreement of the rumor with group biases, attitudes, and prejudices. Put another way, the more anxious a group is, the more a rumor ought to accord with group biases. RE C 138 RO 134 Combining Hypotheses 2 and 3, we put forward the direct determinants of anxiety: importance and reduced sense of control. Simply put, if we have a reduced sense of control over events that are important to us (e.g., job security), then we feel anxious. 132 133 Hypothesis 6. Anxiety predicts agreement of the rumor with group biases, attitudes and prejudices. 160 1.4. Accuracy of rumor 161 Accuracy refers to how veridical a rumor is. Research investigating the overall accuracy of rumors seems contradictory. Allport and Postman concluded: “So great are the distortions . . . that it is never under any circumstances safe to accept rumor as a valid guide for belief or conduct.”46 On the other hand, Davis summarized his research with a seemingly opposite result: “Between 75 and 95% of grapevine information is correct, although most of the stories are incomplete in detail.”47 Some situations apparently yield accurate rumors; others, inaccurate. Given a rumor about an important topic, Buckner proposed that group members not only interact to ensure that they pass and obtain the message accurately; they interact in an attempt to 163 164 165 166 167 168 169 UN 162 CO R 159 158 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 RO 174 175 DP 173 TE 172 RE C 171 distill the facts.48 One member at any one time may compare and contrast discrepant versions of the rumor from different sources and pass along a revised version of the rumor. The revision will be more valid if members are able to pass a more accurate rendition of the rumor, and more distorted if group members are less able to make a valid revision of the rumor. Buckner proposed a set of variables that impact a group’s ability to obtain more accurate revisions of the rumor.49 The variables posited by Buckner pertinent to this investigation include (our terms): group skepticism, established communication channels and rumor activity. Group skepticism refers to the tendency of a group to require credible substantiation before granting credence. Buckner posited that groups that were “gullible” or had “meager standards of evidence” were less able to make accurate revisions to the rumor. Conversely, it follows that groups characterized by skepticism would tend to yield more veridical rumors. Established communication channels refer to stability of communication pathways. Buckner posited that stable communication networks (e.g., an established office grapevine) would yield accurate rumors because persons who communicate with one another over time can judge source credibility; novel or disrupted channels of communication (e.g., as in a natural disaster) afford no such advantage. Regarding rumor activity, Buckner posited that, given some portion of the group that was able to revise the rumor accurately, greater activity would lead to accuracy. Buckner did not give this process a name; we have dubbed this idea the “Percolation Analogy.” Rumors percolate in a group as water circulates in a coffee maker. A group with a skeptical mindset and established communication channels is analogous to good coffee grounds. A group that is gullible and whose communication channels are novel or disrupted is analogous to bad coffee grounds. Rumor activity is analogous to water re-circulating through the grounds. The resulting good cup of coffee is analogous to an accurate rumor; a bad cup of coffee to a distorted rumor. Depending upon the grounds, re-circulation will yield either a good or bad cup of coffee. Depending upon group skepticism and established channels of the group, rumor activity will yield either an accurate or a distorted rumor. These ideas are consistent with heath’s contention “. . . that people discover [information] through their contact with persons inside and outside the company. In this sense, interpretations produce information (even the lack of it) that increases or decreases the amount of uncertainty an interpreter experiences.”50 These considerations lead to CO R 170 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 OF 6 Hypothesis 7. Activity, group skepticism and established communication channels predict accuracy. Based upon the literature, therefore, we propose a model (Hypotheses 1–7). 2. Methods 205 Top-level corporate PR officers and consultants were mail surveyed regarding episodes of harmful or potentially harmful rumors that they had experienced as part of a larger survey on rumor prevalence, effects, and management strategies. Methodology pertinent to the current investigation is reported here. Additional details pertaining to survey sample, instrument development, design and administration, and procedures are reported elsewhere.51 206 207 208 209 UN 204 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 7 2.1. Sample 211 221 Seventy-four completed questionnaires (response rate = 33%) were obtained from an initial sample of 238 individuals comprised mainly of the Arthur W. Page Society. Nine surveys were discarded because most or all of the responses pertinent to this study were blank. Episodes that were gossip were excluded from analysis resulting in the deletion of four surveys. The resulting final sample size was 61. Members of the sample were typically current or former Senior Vice Presidents of Communications at large Fortune-500 corporations (69%) or CEOs of well-established PR consulting firms (26%). Seventy-two percent of the sample was male. Average age of respondents was 51.63 years (SD = 8.76); average experience in PR was 26.23 years (SD = 9.52). Industries in which the sample worked or consulted varied widely. The sample was therefore representative of the highest level and most experienced population of corporate PR and communications personnel available. 222 2.2. Survey instrument 223 We first defined rumor as “. . . an unverified bit of information about something of importance to a group. It is like news in every way except that it is not verified. It may or may not be true. It may be spread by word-of-mouth, fax, electronic mail, or any other communication channel. It is often introduced by the phrase ‘I heard that. . . .”’ Overall rumor prevalence, effects of rumors, the use and effectiveness of specific strategies for preventing and/or neutralizing rumors were first assessed (results are reported elsewhere52 ). Respondents were then directed to think of a particular rumor that had reached their ear that they had considered harmful or potentially harmful. Half of the surveys asked respondents to choose a rumor that had been proven true (and the other half, false) “beyond a reasonable doubt.” They recounted the rumor and described the situation out of which the rumor arose. They were then asked to answer a set of items with respect to one group of people who had heard and/or passed the rumor and for the point in time just prior to when the rumor had been proven true (or false). These items were on a 5-point Likert-type scale and were as follows. 217 218 219 220 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 DP 216 TE 215 RE C 214 2.2.1. Rumor activity Rumor activity was assessed by averaging responses to two questions (coefficient α = .70). First, “When a rumor is passed from one person to another, it may involve a great deal of interaction (discussion and clarification), or it may involve very little interaction. Overall, how interactive was a typical discussion of the rumor?” Second, “Overall, how active (frequently transmitted) was the rumor?” CO R 213 2.2.2. Uncertainty Uncertainty was assessed using the question “Overall, how uncertain (filled with questions about what current event meant or what future events were likely to occur) were people?” 2.2.3. Reduced sense of control Reduced sense of control was assessed using “Overall, to what extent did people have a reduced sense of control or power over events affecting their lives?” Anxiety was assessed UN 212 RO 210 249 250 251 252 253 using “Overall, how anxious (worried/concerned) were people?” Importance was assessed by “Overall, how important (significant, consequential) was the rumor to people?” RO 248 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 OF 8 2.2.4. Belief in rumor Belief in rumor was measured using “Overall, how confident were people that this rumor was true?” Agreement with biases was assessed by “Overall, to what extent did the rumor agree with or was consistent with peoples’ biases (preconceived notions, attitudes, and/or prejudices)?” 262 3. Results 263 Means, standard deviations, and zero-order correlations are presented in Table 1. The set of rumors encountered in this investigation tended to be either true or false (as requested) and important (use of transformed variables did not substantially alter results). All other distributions were roughly “bell-shaped” across the five possible values. All of the 12 hypothesized relationships were in the predicted direction and all but two were significant at the .05 level (one of which was marginally significant at the .10 level). Missing data were then replaced using the mean of all values obtained for each variable. Not more than two values were missing per variable, resulting in the replacement of a total of seven missing values. To test the hypothesized relationships path analysis was performed using the hypothesized model and two mediation analyses were performed on subsets of the model. These analyses were performed using the analysis of moment structures program (AMOS version 4.01). In path analysis, the match between the predicted and obtained covariance matrices is measured via a Chi-square test. Significant Chi-squares show substantial differences between the predicted and observed pattern of covariances. Because Chi-square values are sensitive to sample size (larger samples are likely to show significance) other indices of fit are also used. These include the Tucker–Lewis index (TLI),53 comparative fit index (CFI),54 and the root mean square error of approximation (RMSEA).55 TLI and CFI range between 0 and 1, with values of .90 and above indicating good fit. A value of .05 or less for the RMSEA also indicates good fit. Standardized path coefficients for the hypothesized model and squared multiple correlations (SMCs) for endogenous variables are presented in Fig. 2. Though all paths were in the predicted direction, fit indices did not indicate a good fit (Chi-square(33) = 54.61, p = .01; TLI = .75; 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 TE 259 260 RE C 258 CO R 256 257 UN 255 DP 261 2.2.5. Rumor accuracy Rumor accuracy was measured by “For the rumor, how accurate or true (as opposed to distorted or false) did the rumor prove to be?” The extent to which channels of communication were established (established channels) was measured by “A rumor may pass through an established communication network (e.g., an office grapevine) or through a group of people who never communicated before. Overall, how established were the communication channels of the group?” Group skepticism was assessed by “Overall, to what extent would you characterize the people as gullible (as opposed to skeptical)?” (reverse-scored). 254 UN CO R RE C TE OF DP RO N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 9 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 TE DP RO 10 Fig. 2. Standardized path coefficients and SMCs for hypothesized model. Coefficient typeface indicates significance, bold: p < .05, SMCs are listed atop each endogenous variable. 290 291 292 293 294 RE C 289 CO R 288 CFI = .82; RMSEA = .10). Two path coefficients were especially weak: importance → rumor activity and established channels → rumor accuracy. The mediation role of reduced sense of control in the relationship between uncertainty and anxiety was then investigated separately. Fig. 3 presents the standardized path coefficients and SMCs for the model specifying both direct and indirect paths from uncertainty to anxiety. The direct effect appears to be weak while the indirect paths, strong. A Sobel test56 also indicated that the indirect effect of uncertainty upon anxiety via reduced sense of control was significantly different from zero (Sobel test z = 2.96, p = .003). Reduced sense of control appears to mediate the effect of uncertainty on anxiety. UN 286 287 Fig. 3. Standardized path coefficients and SMCs for model of direct and mediated uncertainty → anxiety effects. Coefficient typeface indicates significance, bold: p < .05, SMCs are listed atop each endogenous variable. OF 11 RO N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 DP Fig. 4. Standardized path coefficients and SMCs for model of direct and mediated importance → rumor activity effects. Coefficient typeface indicates significance, bold: p < .05, SMCs are listed atop each endogenous variable. 309 3.1. Revised model specification 310 In an attempt to improve model fit, model specifications suggested by the data were considered. Two paths were clearly not supported: importance → rumor activity and established channels → accuracy. First, the importance → rumor activity path appears to be mediated via anxiety. We hypothesized this relation. The direct path was therefore deleted from the model. Second, the established channels → accuracy path simply did not appear to be all that important. Given that Buckner had only speculated on this point, we deleted it from the model. Fit indices at this point still indicated a poor fit with the data (Chi-square(26) = 46.38, p = .008; TLI = .76; CFI = .83; RMSEA = .11) despite the fact that all hypothesized paths were significant. Modification indices then suggested two further alterations. First, the path from anxiety to belief was recommended. It is plausible that the more anxious and concerned people are, the more likely they are to believe a rumor.57 Greater anxiety has been found to affect critical ability in rumor episodes.58 This path was therefore added. Modification indices also suggested that we allow uncertainty and importance to be correlated. This would imply that people were filled with questions with regard to important issues. This was likely in this investigation because we had asked participants to recall specific rumors that had been harmful or potentially harmful—and thus, important. Standardized path coefficients and SMCs for this 299 300 301 302 303 304 305 306 307 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 RE C 298 CO R 297 UN 296 TE 308 The mediation role of anxiety in the relationship between importance and rumor activity was also investigated in a separate analysis. Fig. 4 presents the standardized path coefficients and SMCs for the model specifying both direct and indirect paths from importance to activity. Again the direct effect appears to be weaker than the indirect. A Sobel test also indicated that the indirect effect of importance upon rumor activity via anxiety was significantly different from zero (Sobel test z = 2.53, p = .01). Anxiety appears to be a strong mediator of the effect of importance on rumor activity. Thus, support our hypothesized model was mixed. Some parts of the model seem to fit the data better than others. Support for each hypothesis is summarized here: with regard to rumor activity, importance did not emerge as a significant predictor, however, the mediation roles of reduced sense of control and of anxiety were supported (Hypotheses 1–4). With regard to belief in rumor, both Hypotheses 5 and 6 were supported. Finally with regard to rumor accuracy, established channels of communication did not emerge as a significant predictor (Hypothesis 7). 295 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 TE DP RO 12 Fig. 5. Standardized path coefficients and SMCs for final model. Coefficient typeface indicates significance, bold: p < .05, SMCs are listed atop each endogenous variable. 328 329 final model are presented in Fig. 5. Fit indices indicated an acceptable fit between the model and the data (Chi-square(24) = 33.28, p = .10; TLI = .88; CFI = .92; RMSEA = .08). All coefficients were significant. RE C 327 4. Discussion 331 338 This investigation found mixed support for the overall model posited. Given the correlational and interactive nature of path analysis, and given the limitations, we here submit those conclusions in which we have greater confidence: with regard to rumor activity, we found evidence for the role of anxiety, uncertainty and belief. Between these variables, we found evidence for the mediation roles played by reduced sense of control (uncertainty → reduced sense of control → anxiety) and by anxiety (importance → anxiety → rumor activity). With regard to belief in rumor, we found evidence for the role of biases. And with regard to accuracy, we found evidence for the role played by rumor activity and group skepticism. 339 4.1. Limitations 340 We note several limitations of this study. First, data were self-reported and therefore it is possible that responses were subject to several well-known memory biases (e.g., remembering only the most vivid rumor episodes). This is a common and valid criticism of any self-report survey technique. Future research efforts should attempt to incorporate a multi-method approach to data collection. Second, our items attempted to assess average individual or group-level variables (e.g., “how anxious were people [in the group within which the rumor circulated]?”) from individual responses. We have no way of appraising how representative our individual 333 334 335 336 337 341 342 343 344 345 346 UN 332 CO R 330 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 13 370 4.2. Implications for rumor theory 371 Several implications for rumor theory accrue from this research. First, there is evidence for the mediation role of anxiety in the importance → activity relationship in rumors. As discussed in Section 1, we think this relationship more clearly reflects Rosnow’s reconceptualization of importance as “outcome-relevant involvement.”60 Another way of stating this is that rumors about “important” issues (e.g., campus drug use) that, however, do not evoke at least some anxiety (e.g., possibly the rumor that students were caught smoking marijuana during a final exam61 ) are less likely to be transmitted. Put another way, anxiety may account for most of the variance in rumor activity that has often been attributed to importance. Further research should explicitly test for this indirect effect. Second, there is evidence that at least some uncertainty leads to anxiety through a reduced sense of control. This arrangement explicitly bridges two important predictors in rumor transmission. Whatever uncertainty remains after accounting for reduced sense of control is of a less urgent variety—a purely intellectual puzzle—and thus, less likely to lead to transmission. People faced with a puzzle that is not purely intellectual—for example, one on which their livelihood depends—will feel the reduced sense of control, heightened anxiety, and will participate in the rumor mill to solve the puzzle. Given the overall high importance ratings assigned to the rumors in our sample, we would speculate that many of the rumor puzzles we tapped into were of the not-merely-intellectual variety. Future research should explicitly test for this relationship. 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 DP 352 TE 351 RE C 350 CO R 349 UN 348 RO 369 sample’s responses are. For example, our sample may not have a representative sense of how anxious people were, how active the rumor was, nor how strongly people believed the rumor. The data is, in this sense, “secondhand.” Given their leadership position and role in corporate communications and extensive experience, however, we think that of any persons, our sample stood the best chance of making an accurate group assessment. In any event, future research should attempt to use more than one person’s responses per rumor episode. Third, we reiterate that we requested respondents to remember rumors that had proven true or false; because we were interested in accuracy we ignored rumors that were still neither verified nor falsified. The results may therefore only apply to rumors proved true or false and not to those rumor episodes in which the rumor will/does remain unverified/ unfalsified. The fourth limitation involves a perennial trade-off in survey research: response-rate versus reliability assessment. To assess item reliability, redundant items are often included in a questionnaire, yet this often has the effect of annoying participants and reducing response rate.59 In an effort to enhance response rate of this very busy sample of experienced PR practitioners, most measures were single-item and we were therefore unable to assess item reliability. On the positive side, we note that the sample was quite literally the “best of the best.” Despite the limitations, the responses were obtained from extremely competent and experienced women and men who—based upon write-in comments and open-ended responses—took the questionnaire quite seriously. This was not a sample of undergraduate psychology majors. With these considerations in mind, we proffer our conclusions here as suggestive of the way in which aspects of rumor episodes may be related. Future investigations should seek to redress these limitations. 347 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 OF 14 419 4.3. Implications for PR practitioners 420 Several implications for the PR professional are implied by this research for the three outcomes considered in this investigation. Obviously the objectives of the PR professional would be to prevent and/or reduce harmful rumor activity and belief, and to increase rumor accuracy. We consider two of these here. First, to prevent and reduce harmful rumors, PR professionals can organize their efforts as to detect and diminish each of the three direct “ingredients” of rumor activity: uncertainty, anxiety and belief in the rumor. The reduction of uncertainty, of course, is a primary objective for the communications professional.65 Specifically, however, we note the importance of enhancing their formal communications so as to be honest, timely, reliable, and consistent.66 Given an open culture, communications personnel are eager to dispel uncertainty.67 However, in cases where uncertainty cannot or is not permitted to be dispelled, communications 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 421 422 423 424 425 426 427 428 429 430 DP 395 TE 394 RE C 393 CO R 392 UN 391 RO 418 Third, there was evidence that anxiety may affect belief through agreement with group biases. Although this relationship had been described qualitatively in previous literature,62 these are among the first quantitative results in an investigation of rumor to support these ideas. The path from anxiety—through biases—to belief is projective in character. For example, anxious about job security, department members may believe downsizing rumors because these rumors match attitudes of distrust toward the organization.63 Fourth, there is some evidence for the percolation analogy in rumor accuracy, broadly speaking. Rumors that actively “percolate” in a skeptical group tend to become more accurate. These conclusions resonate with Buckner’s conceptualizations that attempted to delineate those situations/factors that enable groups to establish rumor veracity/falsity. Further research should refine and explore group factors associated with accuracy. There is currently a dearth of research in this area; to our knowledge, our investigation represents the initial effort to test these ideas empirically. Fifth, this investigation highlights the need for multi-level approaches to understanding rumor. We note that three of the variables we investigated were attributes of communication networks: rumor activity measured how “interactive” typical rumor discussions among group members were and how much the rumor seemed to “recirculate” among group members, established channels referred to how stable communication pathways in the group were, and rumor accuracy of course referred to how veridical the rumor circulating in the group had become. The remaining variables were aggregated attributes of the individuals; for example, group anxiety was the average sense of anxiety/concern/worry that people felt. These variables were at an individual-level. The combination of levels represents a better understanding of rumor phenomena in which individual attributes interact in the context of social network attributes. Future research should seek to extend this multi-level approach. Finally, this work represents a beginning attempt to conceptualize and integrate related rumor phenomena—activity, belief, and accuracy—into one systematic model. Such efforts are needed because human and communication phenomena are often complex and interactive.64 Rumor is an especially complex phenomenon, involving both group, network, and individuallevel variables, and it has a decidedly social and episodic flavor to it. Future rumor research should attempt to retain the integrative nature of this work. 389 390 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 RO 435 436 DP 434 professionals face a more challenging task. If uncertainty cannot be reduced, our model would direct PR personnel toward enhancing public’s sense of control. PR officers can seek to enhance the group’s sense of control by flavoring and structuring uncertainty.68 Flavoring/structuring uncertainty refers to actions which set boundaries about uncertainty, such as giving a timeline for forthcoming information, discussing the values by which change decisions will be made, and explaining why no comment is being made.69 Along this line, Pinsdorf advises “first and foremost: candor is the best policy. Admitting that you don’t know or are not going to answer a particular question is not as damaging a attempting to cover up or lie.”70 For example, in situations where confirmation of layoff rumors cannot be given, PR officers can discuss how layoffs decisions—should they occur—are likely to be made. Similarly, managers fulfilling a PR function could assist employees in planning for worst-case scenarios and even permit employees to participate in layoff decision-making.71 The point is that knowledge and participation should enhance the group’s sense of control and thus inhibit the uncertainty → anxiety relationship. Second, to reduce belief in a harmful false rumor, PR officials can refute the rumor using appropriate and credible sources,72 succinctly, and in a timely manner.73 Our results support PR academics that advocate for the “take charge” rather than the “sit-on-it” school.74 Also, we note that many organizational rumors seem to find fertile ground for belief in attitudes of mistrust. Persons that are distrustful of the corporation are likely to believe rumors that reflect negatively on the corporation. More broadly, the positive effects of participative cultures and symmetric communication systems have been noted as one characteristic of excellence in PR.75 Indeed, trust is “symmetrical concept.”76 PR officials can specifically seek to increase trust as a way of reducing belief in the rumor.77 This again accords with Pinsdorf’s admonition to conduct business “. . . honestly with an eye to public exposure.” She details how Bank of Boston’s projection of an elitist attitude and failure to quickly admit errors hampered efforts to quell the effects of rumor activity (that the Bank was connected to the mafia). In terms of our model, these actions decreased trust in the organization and made seemingly incredible rumors plausible (biases → belief path). Such trust can go a long way to minimize restrictions that PR officers often face. Former Vice Chairman and CFO for the USG Corporation, Eugene Miller, told the story of a takeover fight in which, though hampered in his communication to employees by SEC guidelines, “happily, we have a very good relationship with our employees . . . and while our communication was limited, we did have their support.”78 TE 433 15 RE C 432 CO R 431 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 Acknowledgments 464 We thank the Institute for Public Relations (IPR) for funding this study and administering survey mailings. We thank Michelle Hinson for flawlessly executing these mailings and Jack Felton for his guidance and assistance throughout this project. We thank Don Bates, Lee Brown, Neil Fagenbaum, Don Hantula, Robert Heath, Jessica Jameson, Bruce MacDonald, Laurie Maynard, Bill McKee, David Neumann, Frank Ovaitt, Marc Renzoni, Ralph Rosnow, Randall Simonetti, and James Tolley for their assistance during the pilot phase of this project. We thank the survey respondents. We thank Ralph L. Rosnow and Bruce Austin for reviewing earlier drafts of this manuscript. 465 466 467 468 469 470 471 UN 463 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 OF 16 RO Nicholas DiFonzo, Ph.D., is an Associate Professor of Psychology at Rochester Institute of Technology in Rochester, NY. He has the specialization in the topic of managing organizational rumors. 472 Prashant Bordia, Ph.D., is Senior Lecturer at the University of Queensland in Brisbane, Australia. He has the specialization in the topic of managing organizational rumors. DP 473 References 475 [1] N. DiFonzo, P. Bordia, How top PR professionals handle hearsay: corporate rumors, their effects, and strategies to manage them, Public Relations Review 26(2000), pp. 173–190. [2] H. Unger, Psst—heard about Pop Rocks? Business rumors and how to counteract them, Canadian Business 6(1979), pp. 39. [3] M.K. Pinsdorf, Communicating When Your Company is Under Siege: Surviving Public Crisis, Lexington Books, Lexington, MA, 1987. [4] P. Bordia, R.L. Rosnow, Rumor rest stops on the information highway: a naturalistic study of transmission patterns in a computer-mediated rumor chain, Human Communication Research 25(1998), pp. 163–179. [5] N. DiFonzo, P. Bordia, A tale of two corporations: managing uncertainty during organizational change, Human Resource Management Journal 37(1998), pp. 295–304; N. DiFonzo, P. Bordia, R.L. Rosnow, Reining in rumors, Organizational Dynamics 23(1994), pp. 47–62. [6] N. DiFonzo, P. Bordia, op. cit., 1998, 2000; M.K. Pinsdorf, op. cit. [7] N. DiFonzo, P. Bordia, op. cit., 2000. [8] A.M. Freedman, Rumor turns fantasy into bad dream, Wall Street Journal (1991) B1, B5. [9] F.W. Koenig, Rumor in the Marketplace: The Social Psychology of Commercial Hearsay, Auburn House Publishing Company, Dover, MA, 1985. [10] N. DiFonzo, P. Bordia, op. cit., 1998. [11] N. DiFonzo, E.E. Toth, Annotated Bibliography of Recent and Significant Psychological Research of Import to Public Relations Practitioners, Institute for Public Relations, Gainesville, FL, 2001. [12] R.L. Rosnow, A.J. Kimmel, Rumors, Encyclopedia of Psychology, Vol. 7, Oxford University Press, Oxford, 2000, pp. 122–123. [13] G.W. Allport, L. Postman, The Psychology of Rumor, Holt, Rinehart & Winston, New York, 1947. [14] M. Gluckman, Gossip and scandal, Current Anthropology 4(1963), pp. 307–316. [15] R.L. Rosnow, Inside rumor: a personal journey, American Psychologist 46(1991), pp. 484–496. [16] C.R. Berger, J.J. Bradac, Language and Social Knowledge: Uncertainty in Interpersonal Relations, Edward Arnold, London, 1982; F.J. Milliken, Three types of perceived uncertainty about the environment: state, effect, and response uncertainty, Academy of Management Review 12(1987), pp. 133–143; R.L. Rosnow, op. cit. [17] G.W. Allport, L. Postman, op. cit.; N. DiFonzo, P. Bordia, op. cit., 1998; D.M. Schweiger, A.S. Denisi, Communication with employees following a merger: a longitudinal field experiment, Academy of Management Journal 34(1991), pp. 110–135. [18] N. DiFonzo, P. Bordia, op. cit., 1998; T. Shibutani, Improvised News: A Sociological Study of Rumor, Bobbs-Merrill, Indianapolis, 1966. [19] R.L. Rosnow, op. cit. [20] R.L. Heath, Management of Corporate Communication: From Interpersonal Contacts to External Affairs, Lawrence Erlbaum Associates, Hillsdale, NJ, 1994, p. 97. 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 RE C 478 479 CO R 477 UN 476 TE 474 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 RO 517 518 DP 516 TE 514 515 RE C 513 17 [21] R.L. Rosnow, op. cit., p. 487. [22] J.L. Esposito, Subjective factors and rumor transmission: a field investigation of the influence of anxiety, importance, and belief on rumormongering, Doctoral dissertation, Dissertation Abstracts International, 48, Temple University, 1986/1987, 596B; R.L. Rosnow, J.L. Esposito, L. Gibney, Factors influencing rumor spreading: replication and extension, Language and Communication 8(1988), pp. 29–42. [23] C.J. Walker, C.A. Beckerle, The effect of anxiety on rumor transmission, Journal of Social Behavior and Personality 2(1987), pp. 353–360. [24] R.L. Rosnow, op. cit. [25] G.W. Allport, L. Postman, op. cit. [26] Ibid, p. 34. [27] R.L. Rosnow, op. cit. [28] K. Back, L. Festinger, B. Hymovitch, H. Kelley, S. Schachter, J. Thibaut, The methodology of studying rumor transmission, Human Relations 3(1950), pp. 307–312. [29] M.E. Jaeger, S. Anthony, R.L. Rosnow, Who hears what from whom and with what effect: a study of rumor, Personality and Social Psychology Bulletin 6(1980), pp. 473–478. [30] J.L. Esposito, op. cit. [31] N. DiFonzo, P. Bordia, R.L. Rosnow, op. cit. [32] R.L. Rosnow, op. cit. [33] R.L. Rosnow, J.H. Yost, J.L. Esposito, Belief in rumor and likelihood of rumor transmission, Language and Communication 8(1986), pp. 29–42. [34] R.L. Rosnow, J.L. Esposito, L. Gibney, op. cit. [35] R.L. Rosnow, J.H. Yost, J.L. Esposito, op. cit.; N. DiFonzo, P. Bordia, R.L. Rosnow, op. cit. [36] R.L. Rosnow, op. cit. [37] S.J. Ashford, Individual strategies for coping with stress during organizational transitions, Journal of Applied Behavioral Science 24(1988), pp. 19–36; S.M. Miller, Predictability and human stress: toward a clarification of evidence and theory, Advances in Experimental Social Psychology 14(1981), pp. 203–256; K.I. Miller, P.R. Monge, Social information and employee anxiety about organizational change, Human Communication Research 11(1985), pp. 365–386; D.M. Schweiger, A.S. Denisi, op. cit. [38] T.L. Albrecht, M.B. Adelman, Communicating Social Support, Sage, Newbury Park, CA, 1987; D.T. Bastien, Common patterns of behavior and communication in corporate mergers and acquisitions, Human Resource Management 26(1987), pp. 17–33; C.R. Berger, J.J. Bradac, op. cit.; D.J. Terry, N.L. Jimmieson, Work control and employee well-being: a decade review, in: C.L. Cooper, I.T. Robertson (Eds.), International Review of Industrial and Organizational Psychology, Wiley, New York, 1999, pp. 95–148; C.J. Walker, B. Blaine, The virulence of dread rumors: a field experiment, Language and Communication 11(1991), pp. 291–297. [39] P. Bordia, E. Hunt, N. Paulsen, D. Tourish, N. DiFonzo, Uncertainty during organization change: it is all about control, 2001, submitted for publication. [40] R. Nisbett, L. Ross, Theory maintenance and theory change, Human Inference: Strategies and Shortcomings of Social Judgment, Prentice-Hall, Englewood Cliffs, NJ, 1980, Chapter 8, pp. 167–192; J.M. Olsen, N.J. Roese, M.P. Zanna, Expectancies, in: E. Tory Higgins, A.W. Kruglanski (Eds.), Social Psychology: Handbook of Basic Principles, Guilford Press, New York, 1996, Chapter 8, pp. 211–238. [41] G.W. Allport, L. Postman, op. cit.; R.H. Knapp, A psychology of rumor, Public Opinion Quarterly 8(1944), pp. 22–27. [42] F.H. Allport, M. Lepkin, Wartime rumors of waste and special privilege: why some people believe them, Journal of Abnormal and Social Psychology 40(1945), pp. 3–36. [43] H.T. Buckner, A theory of rumor transmission, Public Opinion Quarterly 29(1965), pp. 54–70; J.S. Victor, Satanic Panic: The Creation of a Contemporary Legend, Open Court, Chicago, 1993. [44] G.W. Allport, L. Postman, op. cit. [45] M. Rosenthal, Where rumor raged, Trans-Action (1971), pp. 34–43. [46] G.W. Allport, L. Postman, op. cit., p. 148. [47] K. Davis, Human Behavior at Work, McGraw-Hill, San Francisco, 1972, p. 55. [48] H.T. Buckner, op. cit. [49] See also T. Shibutani, op. cit., Chapter 1. CO R 512 UN 511 OF N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 569 [54] [55] 570 571 [56] 572 573 574 [57] 575 576 577 578 579 580 [58] [59] [60] [61] [62] 581 582 [63] 583 584 [64] 585 586 587 588 [65] 589 590 591 [66] [67] 592 593 594 [68] 595 596 597 598 599 [69] [70] [71] [72] 600 601 602 603 604 605 [73] [74] [75] 606 607 608 609 610 611 612 [76] [77] [78] OF RO 568 DP 566 567 R.L. Heath, op. cit., p. 90. N. DiFonzo, P. Bordia, 2000, op. cit. Ibid. P.M. Bentler, D.G. Bonett, Significance tests and goodness of fit in the analysis of covariance structures, Psychological Bulletin 88(1980), pp. 588–606. P.M. Bentler, Comparative fit indexes in structural models, Psychological Bulletin 107(1990), pp. 238–246. M.W. Browne, R. Cudeck, Alternative ways of assessing model fit, in: K.A. Bollen, J.S. Long (Eds.), Testing Structural Equation Models, Sage, Newbury Park, CA, 1993, pp. 136–162. R.M. Baron, D.A. Kenny, The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations, Journal of Personality and Social Psychology 51(1986), pp. 1173–1182. J. Prasad, The psychology of rumour: a study relating to the great Indian earthquake of 1934, British Journal of Psychology 41(1935), pp. 129–144. R.L. Rosnow, J.L. Esposito, L. Gibney, op. cit. D.A. Dillman, Mail and Telephone Surveys: The Total Design Method, Wiley, New York, 1978. R.L. Rosnow, op. cit. M. Jaeger, et al., op. cit. G.W. Allport, L. Postman, op. cit.; R. Knapp, op. cit.; P.A. Turner, I Heard it Through the Grapevine: Rumor in the African American Culture, University of CA Press, Berkely, CA, 1993. R.S. Burt, M. Knez, Trust and third-party gossip, in: R.M. Kramer, T.R. Tyler (Eds.), Trust in Organizations: Frontiers of Theory and Research, Sage, Thousand Oaks, CA, 1996, pp. 68–89. K. Miller, Quantitative research methods, in: F.M. Jablin, L.L. Putnam (Eds.), The New Handbook of Organizational Communication: Advances in Theory, Research, and Methods, Sage, Thousand Oaks, CA, 2001, pp. 137–160; R.R. Vallacher, A. Nowak (Eds.), Dynamical Systems in Social Psychology, Academic Press, San Diego, 1994. D.M. Dozier, L.A. Grunig, J.E. Grunig, Manager’s Guide to Excellence in Public Relations and Communication Management, Lawrence Erlbaum Associates, Mahwah, NJ, 1995; R.L. Heath, op. cit. N. DiFonzo, P. Bordia, 1998, op. cit. J.E. Grunig, Symmetrical systems of internal communication, in: J.E. Grunig, D.M. Dozier, W.P. Ehling, L.A. Grunig, F.C. Repper, J. White (Eds.), Excellence in Public Relations and Communication Management, Lawrence Erlbaum Associates, Hillsdale, NJ, 1992, pp. 531–575. N. DiFonzo, P. Bordia, 1998, op. cit.; L. Hirschhorn, Cutting Back: Retrenchment and Redevelopment of Human and Community Services, Jossey-Bass, San Francisco, 1983. N. DiFonzo, P. Bordia, op. cit., 2000. M.K. Pinsdorf, op. cit., p. 113. L. Hirschhorn, op. cit. P. Bordia, N. DiFonzo, V. Travers, Denying rumors of organizational change: a higher source is not always better, Communications Research Reports 15(1998), pp. 189–198; P. Bordia, N. DiFonzo, C.A. Schultz, Source characteristics in denying rumors of organizational closure: honesty is the best policy, Journal of Applied Social Psychology 30(2000), pp. 2309–2321. N. DiFonzo, P. Bordia, R.L. Rosnow, op. cit. M.K. Pinsdorf, op. cit., p. 112. D.M. Dozier, et al., op. cit.; J.E. Grunig, What is excellence in management? in: J.E. Grunig, D.M. Dozier, W.P. Ehling, L.A. Grunig, F.C. Repper, J. White (Eds.), Excellence in Public Relations and Communication Management, Lawrence Erlbaum Associates, Hillsdale, NJ, 1992, pp. 219–250. J.E. Grunig, Symmetrical systems of internal communication, op. cit., p. 558. N. DiFonzo, P. Bordia, 2000, op. cit. E. Miller, Representing the corporation to its owners under fire, in: C.L. Claywood, R.P. Ewing (Eds.), The Handbook of Communications in Corporate Restructuring and Takeovers, Prentice Hall, Englewood Cliffs, NJ, 1989, p. 84. TE 565 [50] [51] [52] [53] RE C 564 CO R 563 N. DiFonzo, P. Bordia / Public Relations Review 150 (2002) 1–18 UN 18