Hesitation to Share Bad News

Abstract

Research on bad news delivery reveals a reliable temporal delay in the onset of the bad news message from the sender to the receiver. Two experiments utilized a false feedback test design to determine whether the delay is better accounted for by negative verbal message planning, politeness, or both. Both studies (Ns = 135 and 138) featured participant-senders who delivered either scripted or unscripted good, neutral, or bad news to a stranger. News valence, delay before response, and reluctance were measured. Both experiments supported the functional politeness explanation. Study 2 also supported the negative verbal message–planning explanation. Implications and limitations are discussed.

Keywords

MUM effect bad news reluctance cognitive biases interpersonal communication negative feedback

People hesitate to share bad news. This tendency is known as a MUM effect (keeping “Mum” about Undesirable Messages; Rosen & Tesser, 1970; 1972; Tesser & Rosen, 1975), and MUM effects have been observed in various contexts and settings (Bisel, Kelley, Ploeger, & Messersmith, 2011; Bisel, Messersmith, & Kelley, 2012; Fisher, 1979; Ploeger, Kelley, & Bisel, 2011; Sussman & Sproull, 1999; Weenig, Wilke, & ter Mors, 2011). For example, the recent popular movies Up in the Air and 50/50 dramatize the difficulties associated with bad news delivery in cases of employment termination and physician-patient communication, respectively. A common behavioral MUM effect is a temporal delay before the onset of the sender’s bad news message (Bond & Anderson, 1987; Dibble & Levine, 2010; Tesser, Rosen, & Tesser, 1971; Uysal & Oner-Ozkan, 2007; Yariv, 2006).

Although this temporal delay has been well demonstrated, the mechanisms producing the effect are less well understood. On one hand, the delay might be nothing more than an artifact of time associated with planning the bad news message. Perhaps senders need additional time to choose their words. Literature on cognitive negativity biases suggests that negative information is perceived to be “heavier on the brain” than positive information of equal extremity (e.g., Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Ito, Larsen, Smith, & Cacioppo, 1998; Taylor, 1991). Choosing words for a bad news message takes longer than choosing words for a good news message because the news is negative and that creates an inherently greater cognitive load.

On the other hand, one can view the delay using functionality and (multiple) goal attainment (Dillard, Segrin, & Harden, 1989) perspectives. For example, senders may delay the onset of their bad tidings in the service of politeness (Brown & Levinson, 1987; Uysal & Oner-Ozkan, 2007), to “fire a warning shot” (Ptacek, Ptacek, & Ellison, 2001), out of consideration for self-presentation (Goffman, 1967), or to soften the impact of the bad news (Brown & Levinson, 1987). Thus, it may be true that bad news presents the news bearer with a more goal-complex, cognitively challenging communicative task, but increased delivery time might also reflect a strategic move in service of communication goals. That is, by hesitating to convey bad news, a sender can portray the self as empathic and show sympathy and sensitivity to the message recipient. Consistent with a politeness perspective, Bond and Anderson (1987) found that negative feedback was delayed longer than positive feedback only when the sender was visible to the recipient. This finding pointed to the delay as being a strategic interpersonal communication display rather than, or in addition to, an intrapersonal artifact of cognition valence biases.

Determining whether the delay is associated with verbal message planning can be accomplished by comparing senders who deliver scripted messages with senders who deliver unscripted messages. To the extent that the delay reflects choosing one’s words and nothing else, then scripting the bad news message should attenuate the delay. To observe a temporal delay despite the delivery of a scripted message would suggest that the delay is not merely a by-product of intrapersonal cognition valence and that the delay might reflect a more functional nature, perhaps associated with politeness.

The current article presents the results of two experiments designed to test whether the delay indeed represents intrapersonal verbal message planning exclusively or if this finding might be explained by something else (e.g., an interpersonal politeness stimulus). Addressing this question benefits bad news delivery research in both theoretical and practical ways. Theoretically, to the extent that the findings clarify the underlying processes producing MUM effects, the explanatory function of theory is enhanced. We believe that bad news research is in dire need of theoretical development on which to scaffold its myriad MUM effect demonstrations. The current research begins by applying politeness theory to move beyond demonstrating the existence and prevalence of MUM effects and addresses the question of whether certain MUM effects serve a communication purpose. On a practical level, this research can be utilized to further identify strategies people enact when faced with delivering bad news, which facilitates the development of communication interventions. Knowledge of this type applies to physicians, supervisors, police, clergy, and others who routinely deliver bad news. For example, physicians are commonly taught to deliver bad news quicker by planning ahead what they will say to their patient (Eggly et al., 2006). Should the delay come to represent something other than just choosing one’s words, such advice would deny physicians what could turn out to be an important communication tool.

This article proceeds as follows. First, bad news, research on negativity biases, and politeness theory are reviewed. Next, arguments are presented for the research propositions. Two experiments are then reported wherein news valence and message scriptedness are induced within the context of a live interaction. Finally, limitations as well as theoretical, practical, and research implications are discussed.

Bad News and Mere Cognitive Negativity

The current study adopts Dibble and Levine’s (2010) conceptualization of bad news as a message communicating information that is previously unknown to the receiver, is anticipated to be personally relevant to the receiver, and is perceived by the delivery agent to be negatively valenced by the receiver. Bad news communicates negative information by definition. Research points to the existence of a pervasive and psychological negativity bias characterized by negative information being weighted heavier than positive information (Rozin & Royzman, 2001; Taylor, 1991). For example, evidence reviewed by Baumeister et al. (2001) showed that negative emotions and negative feedback have more psychological impact than positive emotions or positive feedback. The folk axiom “Losses loom larger than gains” reflects the human tendency to perceptually distort negative information such that it seems more extreme than it actually is (Ito et al., 1998). Moreover, many researchers now consider positive and negative affect to be qualitatively distinct phenomena as opposed to mere end points on the same continuum (e.g., Berscheid, 1983; Diener & Emmons, 1985; Ito et al., 1998).

Many of the outcomes associated with differential processing of negative versus positive information implicate temporality. For example, people spend more time looking at negative information than they do positive or neutral information (Fiske, 1980). Moreover, negative events lead to more cognitive work and more complex cognitive processing than do positive events (Peeters & Czapinski, 1990). Likewise, people experiencing negative emotions tend to gather more information, engage in more cognitively complex processing strategies (i.e., relying less on time-saving heuristics), and employ more effortful systematic elaboration of messages (Taylor, 1991). It is difficult to imagine a scenario wherein additional cognitive work and more complex processing would mean less time spent processing. Thus, the literature on negativity biases suggests that negative events and negative emotions take longer to process than do their positive counterparts.

Although this literature primarily focused on negative biases in recipients of information and not specifically senders, at least two reasons suggest that the cognitive machinery responsible for invoking negative biases might still operate within senders who anticipate delivering negative information to a receiver (although not necessarily to the same degree as experienced by receivers). First, emotional contagion (Hatfield, Cacioppo, & Rapson, 1992) or empathic concern might be enough to trigger negative mental states in senders and thus activate the associated biases. Second, research already documents considerable costs personally incurred by senders of bad news to the senders’ reputation and/or competence (e.g., Buckman, 1984; Weenig et al., 2011), and these costs attest to the negativity experienced by senders. If sharing bad news was in no way negative for senders, then there would be no reason for senders to report feeling uneasy, reluctant, and hesitant to share bad news. Indeed, these costs might very well represent the essence of the MUM effect (Tesser & Rosen, 1975). Thus, it is our position that having to deliver bad news represents a negative event for senders of that news and that this negative event can trigger processing in ways that are different for senders delivering good news.

Because cognitive negativity biases can be expected in senders faced with breaking bad news, an argument can be made as follows. Negative events are those that potentially or actually create adverse outcomes for an individual (Taylor, 1991). The costs incurred by senders suggest the adverse nature of sharing bad news; thus, bad news sharing can be considered a negative event for senders. If bad news sharing is negative for senders, and if negative events in general take longer to deal with than do positive or neutral events, then a delay in the transmission of bad news (relative to good news) might be observed for no other reason than that having to plan a negative verbal message would require additional time than planning an otherwise equivalent positive verbal message.

Bad News and Politeness Theory

Because bad news delivery involves self and other’s face, face threats, and facework, politeness theory (Brown & Levinson, 1987) is relevant. Politeness theory holds that individuals are motivated to reduce threats to two types of face: positive face and negative face. Positive face refers to a person’s desire to be approved of in social interaction (Goffman, 1967). Negative face is a person’s desire to remain free from imposition (Brown & Levinson, 1987).

Politeness theory further assumes that as one interacts with others, one rationally estimates the face-threatening potential of any actions he or she is about to take, then selects a communication strategy that maximizes the positive faces of the self and the other, while minimizing transgressions against the negative faces of self and other. The practice of doing facework in this manner is termed politeness. Moreover, politeness strategies are not limited to the linguistic realm. That is, people routinely enact nonlinguistic means to communicate mental states such as liking/disliking, comfort/discomfort, as well as politeness-related concerns for self/other, and so on (e.g., Mehrabian, 1966). According to politeness theory, the amount of politeness a person enacts is positively and linearly related to the magnitude of the anticipated face threat. This point is particularly relevant to the situation of delivering bad news, and it renders straightforward the application of politeness theory to bad news delivery. By definition, interpersonally transmitted bad news connotes negative consequences for a receiver and is conveyed by a sender (who is at least acting under the perception that the news he or she bears implies something negative for the receiver). Those negative consequences have already been shown to trigger what communication researchers will recognize as positive and negative face concerns on the part of senders who fear being blamed for the news, being put in a negative mood, causing pain to the receiver, and so on (e.g., Buckman, 1984; Tesser & Rosen, 1975). Thus, bad news is by definition and extension face threatening. Furthermore, because the negativity of bad news occurs in gradations (Dibble & Levine, 2010; Fallowfield & Jenkins, 2004), the face threat potential of the news should covary with the negativity of the news. As a result, senders faced with sharing bad news can be expected to vary their politeness maneuvers concomitantly with the anticipated threats to their own and the receiver’s faces.

Consistent with politeness theory, the reluctance experienced by bearers of bad news prior to and while discharging their duties as bad news messengers could be a function of psychological costs associated with managing face. In other words, delaying the onset of the bad news might constitute a conscious or unconscious nonlinguistic face-saving tactic on the part of senders (Bond & Anderson, 1987). The politeness explanation suggests that the delay is not only a function of increased cognitive effort in managing multiple goals when generating what to say but also a strategic delay as a way to mitigate face threat. Enacting reluctance communicates concern for the other and shows the self to be a considerate individual.

One final caveat deserves mention. In practice, it is highly likely that any specific bad news–sharing episode will trigger multiple and simultaneous face threats to both the sender’s and the receiver’s faces. Because the current research does not explicitly focus on the measurement of multiple face threats, a discussion of politeness theory (Brown & Levinson, 1987) serves our need to illustrate how the MUM effect might serve a politeness function. However, researchers who are explicitly interested in documenting the imagined or actual face threats associated with sharing bad news will likely find more comprehensive theoretical satisfaction by leveling identity implications theory (Wilson, Aleman, & Leatham, 1998; Wilson, Kim, & Meischke, 1991). Identity implications theory adapts politeness theory by accounting for multiple threats to both the sender’s and receiver’s faces.

Research Propositions

One type of MUM effect that routinely manifests is a temporal delay before a sender begins to communicate the bad news to the recipient (Bond & Anderson, 1987; Dibble & Levine, 2012, 2010; Uysal & Oner-Ozkan, 2007). Because MUM effects by way of delays and other forms have been well documented since the 1970s (e.g., Bisel et al., 2011; Bisel et al., 2012; Bond & Anderson, 1987; Dibble & Levine, 2010; Tesser & Rosen, 1975; Uysal & Oner-Ozkan, 2007), no additional argument is made beyond the reliable empirical findings obtained in previous research to buttress our expectation that MUM effects will replicate here. Two hypotheses, each addressing a different dependent variable, predict replications of MUM effects. The first hypothesis predicts a MUM effect via senders physically delaying the onset of bad news messages. The second hypothesis predicts a MUM effect in the form of felt psychological reluctance being greater under conditions of bad news than under conditions of good news.

Hypothesis 1 (H1): Bad news will be delayed longer than good news.

Hypothesis 2 (H2): Senders of bad news will report greater felt reluctance than senders of good news.

The primary research objective concerns the nature of the delay before feedback. If the delay reflects additional time associated with planning a negative verbal message and nothing else, then, all else being equal, providing a situation where senders do not have to choose their words should hasten message delivery and negate the delay. Combining good and bad news valence conditions with scripted and unscripted feedback conditions permits a test of this possibility. In this situation the scripted/nonscripted nature of the message would be relevant. Thus, within the bad news condition, senders delivering unscripted bad news should delay their feedback longer than senders who deliver scripted bad news.

Hypothesis 3 (H3): Senders delivering unscripted bad news will delay their response longer than senders who deliver scripted bad news.

On the other hand, the delay could be functional, perhaps to mitigate anticipated face threats. Such an expectation is consistent with politeness theory (Brown & Levinson, 1987). If the delay serves a politeness function, this MUM effect should linger regardless of whether the message is scripted or not. This is because senders are delaying the onset of their scripted bad news as a means by which to foreshadow the message, communicate politeness, and forestall face threats, not necessarily because they require additional time to plan their negative messages. The observation that scripted bad news is shared slower than scripted good news would be consistent with the delay as being politeness related.

Hypothesis 4 (H4): Senders delivering scripted bad news will delay the news longer than senders delivering scripted good news.

We note that H3 and H4 reflecting the negative verbal message planning and politeness explanations are not mutually exclusive. That is, the data may well be consistent with both hypotheses and both explanations. The politeness view does not preclude increased cognitive effort or vice versa. It will be recalled that the goal of this study is to probe the feasibility of a functional explanation and explain variance in the delay to share bad news. Thus, both may co-occur and explain their own shares of variance, and this notion is reflected in the research design and hypotheses.

Study 1

Method

Two experiments were conducted wherein news valence was induced as a continuous independent variable. News valence (i.e., good news, neutral news, bad news) was fully crossed with scriptedness condition (i.e., scripted feedback, unscripted feedback). Time to response, perceived news valence, and self-reported reluctance served as the primary dependent variables. The basic procedural paradigm involved senders relaying false test feedback via face-to-face talking to receivers in the context of a live interaction. This research was internal review board approved.

Participants

Participants (N = 270; 54. 8% female, n = 148; M_age = 22.08 years, SD = 4.70, range = 18-49) were recruited from various communication courses at a large Pacific university. The race-ethnic makeup of the participants was predominantly Asian (51.5%), followed by European American (19.7%), Pacific Islander (9.8%), Latino/Latina (4.5%), and African American (2.3%). Approximately 12.2% of the participants indicated they were multiracial or of some other race-ethnicity.

Participants were paired to create 135 dyads. The sex composition of each dyad was also recorded. Here, 38 dyads featured female scorers and test takers, 28 dyads featured female scorers and male test takers, 44 dyads had male scorers and female test takers, and 25 dyads had male scorers sharing with male test takers.

Procedures

For each session, two participants arrived at the lab.¹ Informed consent was properly obtained. Roles (i.e., test taker, sender) and scriptedness conditions (i.e., scripted, unscripted) were randomly assigned, and test takers would complete a bogus social perception test that presumably held implications about whether or not he or she was socially awkward. Test takers took the “test”; senders shared the results with the test taker. News valence was induced as a “percentage” score on the bogus test (a notebook computer was configured to randomly produce an integer between 10 and 90 inclusive).²

Once the test taker finished, he or she walked his or her score sheet to the sender, who was seated at a table behind the computer that presumably performed the scoring calculations. Senders would key in the test taker’s answers, then activate the “score screen” in the computer before revealing the results. Three video cameras recorded the participants’ behaviors during the entire interaction, and the cameras were operated remotely from a separate nearby control room.

Senders in the unscripted condition were free to reveal the test taker’s results in any way they wished. However, senders relaying scripted feedback did so by saying the following message, which was printed on a laminated sheet:

You scored (insert score here) percent. Scores above 64 percent indicate superior performance. Scores ranging from 37 to 63 percent indicate average performance. And scores below 36 percent indicate inferior performance. Please return to your testing area now, and the experimenter will be with you shortly.

Subsequent review of the video-recorded interactions revealed that three of the scripted participants deviated somewhat from the scripted message they were instructed to deliver.³

After senders finished sharing the test results, both participants completed a questionnaire containing several of the dependent measures. Participants were probed to discover whether they had guessed the study’s true purpose (none indicated they had), fully debriefed, asked not to discuss the experiment with others, thanked for their participation, and dismissed.

Measures

Perceived news valence

All measures are available from the lead author. Perceived news valence was measured to determine whether the various test scores appeared to induce various news valences. Six items developed by the authors assessed perceived news valence. Sample stems included, “The test taker performed well,” “The test taker performed poorly” (reverse scored), and “The test taker likely viewed my feedback as bad news” (reverse scored). Respondents indicated their response using a 7-point, Likert-type scale (1 = strongly disagree, 7 = strongly agree). Higher numbers corresponded to more positive valence judgments. Although test takers’ perceptions were measured, only the senders’ responses are reported here (n = 134). Interitem correlations were high, ranging from .60 to .89 (p < .05 for all). Principal axis factor analysis for this six-item measure revealed a single factor that explained 75.43% of the variance. Cronbach’s alpha was .95 (n = 134).

Time to response

Time to response served as a behavioral indicator. This variable was derived using the elapsed time in seconds as it was displayed by the internal time stamp system within the video-recording equipment. The point at which the sender activates the computer’s “score screen” served as the start point for the timer. The stop-time point was the initial onset of the sender’s verbal message to the test taker. Simple subtraction calculated the length of the delay.

Reluctance

Five items developed by the authors assessed sender reluctance: Stems included “I didn’t want to share my partner’s score with them,” “I was eager to speak to my partner” (reverse scored), “I felt reluctant to tell my partner how they did,” “I wasn’t looking forward to giving my partner their result,” and “I felt no hesitation to share my partner’s score with them” (reverse scored). Senders responded using a 7-point Likert-type scale (1 = strongly disagree, 7 = strongly agree). Higher numbers corresponded to greater felt reluctance. These five items formed a unidimensional scale that accounted for 48.85% of the variance. Cronbach’s alpha for the scale was .82 (n = 132).

Results

Induction Checks

Cases missing data were excluded pairwise, producing slight variations in sample size. Actual Ns are reported by analysis. Table 1 lists the means and standard deviations for all three dependent variables as a function of test score group (i.e., superior, average, inferior). An induction check of news valence was performed to verify the relative valences senders perceived for high, average, and low scores. Comparing the three score groups (from this point forward termed valence categories) revealed a significant omnibus test, F(2, 131) = 151.74, p < .01, η² = .70. Dunnet’s (1980) T3 (pairwise comparison test based on the Studentized maximum modulus) post hoc comparisons, which do not assume equal variances, revealed each mean differed from the other two, p < .01. Inspection of the means reveals that scorers viewed superior scores to be associated with the most positive valences, followed by average scores, with inferior scores being associated with the most negative valences. Thus the induction of news valence was effective and substantial.

Table 1.

Means and Standard Deviations for All Dependent Variables as a Function of Valence Category.

	Study 1			Study 2
Valence category (score range)	M	SD	n	M	SD	n
News valence
Good news (90-64)	6.45	0.51	52	6.44	0.77	41
Neutral news (63-37)	4.44	1.17	35	4.72	1.11	48
Bad news (36-10)	2.00	0.75	47	2.66	1.08	45
Time to response (in seconds)
Good news (90-64)	2.85	1.64	49	2.77	1.82	37
Neutral news (63-37)	4.61_a	4.81	33	4.24_b	2.86	45
Bad news (36-10)	4.57_a	2.18	42	5.13_b	3.20	45
Reluctance
Good news (90-64)	1.87	.80	52	1.92	0.80	40
Neutral news (63-37)	3.44	1.53	35	2.80	1.14	47
Bad news (36-10)	4.78	1.43	47	4.26	1.43	45

Note. For news valence and reluctance, response set is a 7-point Likert-type scale where higher numbers correspond to more positive valence and greater felt reluctance, respectively. Post hoc comparisons using Dunnett’s (1980) T3 (which assumes nonequal variances) were used. For news valence, each mean differs from the others at p < .01. For time to response, within each study, means sharing a subscript do not differ from each other at p < .05 (the different subscripts across studies are for labeling only; statistical comparisons do not cross studies). For reluctance, each mean differs from the others at p < .05.

Sex Effects

Each dependent variable was analyzed for differences with respect to the sex makeup of the dyad. This was done to control for any effects of sex makeup on the process of delivering the news. No significant differences were found for news valence, F(3, 130) = 1.37, p = .25, η² = .03; time to response, F(3, 123) = 0.68, p = .56, η² = .01; or reluctance, F(3, 128) = 1.37, p = .25, η² = .03. The sex of the sender and test taker appeared to make no difference, nor did the sex makeup of the dyad. Thus, the entire sample of senders was collapsed, and the tests of the hypotheses proceeded without regard to the sex composition of the dyads.

Tests of Hypotheses

Table 2 lists the zero-order correlations among test score, news valence, time to response, and reluctance. To test H1 and H2, a 3 × 2 analysis of variance (ANOVA) was performed by fully crossing three levels of news valence (superior, average, inferior) with scriptedness condition (scripted, unscripted). Planned contrasts were then conducted to test H3 and H4.

Table 2.

Zero-Order Correlations Among Primary Variables.

Variable	1	2	3	4
1. Test score	—	.91**	−.25**	−.73**
2. News valence	.81**	—	−.27**	−.78**
3. Delay before feedback	−.30**	−.24**	—	.33**
4. Reluctance	−.62**	−.66**	.17*	—

Note. Coefficients below the diagonal are from Study 1; coefficients above are from Study 2. Ns ranged from 127 to 138. Cases missing data were excluded pairwise.

p = .06, two-tailed. **p < .01, two-tailed.

H1 and H2 predicted replications of the MUM effect, using time to response and reluctance as dependent variables, respectively (see Table 3). The 3 × 2 ANOVA revealed a main effect for valence category on time to response, F(2, 121) = 7.95, p < .01, η² = .11. As can be viewed in Table 1, post hoc probing revealed that inferior scores (i.e., bad news) were delayed significantly longer than superior scores (i.e., good news), p < .05. Using reluctance as the dependent variable revealed an even larger main effect for valence category, F(2, 126) = 43.23, p < .01, η² = .40. Post hoc probing shows that greater reluctance was felt when the news was bad than when the news was good, p < .05 (see Table 1). Thus, the data are consistent with both H1 and H2 in that the MUM effects replicated.

Table 3.

Analysis of Variance Data.

Study 1
Source	df	SS	MS	F	p	η²
Dependent variable = Time to response
Valence category	2	118.14	59.07	7.95	.01	.11
Scriptedness	1	19.447	19.45	2.62	.11	.02
Valence category × scriptedness	2	9.19	4.60	0.62	.54	.01
Error	121	898.84	7.43
Dependent variable = Reluctance
Valence category	2	118.80	59.40	43.23	<.01	.40
Scriptedness	1	0.33	0.33	0.24	.63	<.01
Valence category × scriptedness	2	1.88	0.94	0.68	.51	.01
Error	126	173.11	1.37
Study 2
Source	df	SS	MS	F	p	η²
Dependent variable = Time to response
Sex makeup	1	39.22	39.22	4.57	.04	.04
Valence category	2	71.32	35.66	4.16	.02	.07
Scriptedness	1	34.84	34.84	4.06	.05	.03
Valence category × scriptedness	2	5.53	2.76	0.32	.73	.01
Error	111	951.78	8.58
Dependent variable = Reluctance
Valence category	2	178.23	89.12	55.39	<.01	.49
Scriptedness	1	0.54	0.54	0.33	.57	<.01
Valence category × scriptedness	2	2.37	1.19	0.74	.48	.01
Error	113	181.81	1.61

Note. df = degrees of freedom; SS = sum of squares; MS = mean square.

H3 and H4 test two explanations for the delay in news sharing, negative verbal message planning and politeness, respectively. Table 4 lists the means and standard deviations for both time to response and reluctance as a function of news valence and scriptedness conditions. Pairwise comparisons of cell means from the 3 × 2 ANOVA served to test H3 and H4. Using time to response as the dependent variable, no significant difference emerged between unscripted bad news (n = 23) and scripted bad news (n = 21), t(39.32) = 1.42, p = .17, η² = .05. Thus, the data were not consistent with the mere negative message–planning explanation (H3). However, a planned contrast revealed that scripted bad news (n = 21) was delayed longer than scripted good news (n = 20), t(31.95) = −3.54 p < .01, η² = .28.⁴ Thus, the data were consistent with the politeness explanation (H4).

Table 4.

Means and Standard Deviations by Valence Category and Scriptedness Condition.

	Study 1			Study 2
	GN	NN	BN	GN	NN	BN
Dependent variable = Time to response (in seconds)
Scripted	2.81 (1.96)	4.68 (2.89)	5.90 (3.45)	2.72 (0.98)	4.25 (4.24)	3.65 (1.21)
Unscripted	2.72 (1.69)	3.85 (2.84)	4.46 (2.87)	2.99 (2.14)	5.03 (5.74)	5.28 (2.56)
Dependent variable = Reluctance
Scripted	1.70 (0.68)	2.85 (1.34)	4.28 (1.49)	1.95 (0.89)	3.13 (1.21)	4.61 (1.48)
Unscripted	2.15 (0.87)	2.74 (0.96)	4.25 (1.41)	1.72 (0.63)	3.58 (1.85)	4.81 (1.55)

Note. GN = good news, NN = neutral news, BN = bad news. N = 135 dyads (Study 1), 138 dyads (Study 2). Standard deviations are given in parentheses.

Using self-reported reluctance as the dependent variable produced corroborating results. Reluctance was virtually unchanged whether bad news was scripted or unscripted, t(39.48) = −0.08, p = .93, η² < .01. However, scripted bad news prompted greater reluctance than scripted good news, t(26.28) = −7.05, p < .01, η² = .65. To summarize, whether the feedback was scripted or not, bad news was delayed longer and prompted greater reluctance than good news. Scripting the sender’s message did not significantly attenuate either MUM effect. Thus, the data were more consistent with the functional explanation (H4) than the mere negative message–planning explanation (H3).⁵

Discussion

The goal of Study 1 was to initiate basic research that goes beyond MUM effect demonstrations to uncover the reasons why MUM effects occur. The form of MUM effect under present consideration was the physical delay senders often invoke when faced with delivering a bad news message. It was hypothesized that this delay might emerge due to (1) increased time needed to plan a negative verbal message as an incidental consequence of the tendency for people to weigh negative information heavier than positive information. Alternatively or concurrently, it was hypothesized that the delay might reflect (2) a communication stimulus that signals politeness, communicates sensitivity and empathy, and/or somehow softens the blow of the bad news. The data from Study 1 were more consistent with the politeness explanation.

Although the results from Study 1 suggested a view of the delay as serving a politeness function, a number of design features could limit interpretation of the results. First, it is always a question of whether good and bad news were really instantiated in the minds of the participants. Perhaps social awkwardness is not serious enough to raise the stakes of importance of scoring highly on the test. Furthermore, although consistent with previous research (e.g., Dibble & Levine, 2010), utilizing social awkwardness might present a confound such that senders who delay their feedback do so not because of anticipated bad news but because they might disprefer to interact with a “socially awkward” person. Thus, it would be preferable to create a context that raises the stakes of the encounter and avoids a socially oriented variable.

Next, the induction of the scriptedness conditions could be strengthened. Study 1 senders in the scripted condition were instructed to read their script from a laminated sheet that was placed beside the notebook computer. Once they activated the score screen, senders had to look down, locate the laminated sheet, then deliver their scripted feedback. Thus, in addition to delays purely associated with the news valence, the delay measure would have accidentally captured whatever additional time senders needed to locate the sheet containing their script. To the extent that senders used this “script-locating time” as an excuse to stall their tidings, such a finding would be theoretically meaningful. Unfortunately, the current design precluded any speculation as to this possibility. Thus, a design that circumvents the necessity for senders to look away from the computer in order to locate their scripting information would be desirable. Because these limitations cannot be overcome using the current study, and because replication is always a boon, a second study is necessary.

Study 2

Method

Study 2 was performed to replicate Study 1, using an enhanced design that overcame the potential limitations of Study 1. To this end, two major changes were made. First, the social awkwardness test was replaced by a test of the taker’s intelligence (that we called the Personal Intelligence Assessment). Because high intelligence is widely understood to be preferable, we hoped that this would disambiguate the importance of scoring highly in the sender’s mind and thereby boost the stakes of the good/bad news encounter. For ethical reasons, we did not wish naive test takers to be led to believe, even falsely, anything detrimental about their own intelligence. Therefore, Study 2 sessions also employed a confederate (blind to the hypotheses but who was aware that the test was bogus) who would always act in the role of test taker. In this way, naive participants served as scorer senders and never had to receive news regarding their intelligence.

The second major change was to modify the scoring screen such that it now displayed the sender’s script in addition to the score. In fact, the on-screen message was further configured to display the score as part of the message. With this change, we hoped to improve the delay measure by minimizing the possibility that the sender was hesitating only because he or she could not locate his or her script. Thus, if senders stalled, waited, shuffled around pretending to find their instructions, and so on, it would be difficult to argue that these delays were because they were unaware of what to say. Senders in the unscripted condition would continue to deliver their feedback as before.

Researcher and Confederate Training

Two different confederates were utilized along with nine different experimenters, all blind to the study’s hypotheses. Once the experimental protocol (available from the lead author) had been developed and internal review board approval secured, all researchers (confederates and experimenters) received training on the proper operation of all lab equipment. Next, rehearsal sessions were conducted by combining one researcher to serve as the experimenter, another researcher to dummy the participant role, and one to dummy the role of the full-time confederates. Rehearsal sessions were held until all researchers felt comfortable in their roles and could execute the sessions uniformly. Training could be done rather swiftly and efficiently because experimenters would carry throughout every session a clipboard containing a detailed script and confederates only had two lines of dialog to memorize. All experimenters and confederates received at least five complete rehearsal sessions before they operated using live participants.

Participants and Procedure

Participants (N = 138; 59% male, n = 79; M_age = 20.97 years, SD = 2.90, range = 17-37) were recruited from various communication classes at the same large Pacific university as those from Study 1. (Precautions were taken to ensure that no participant from Study 1 participated in Study 2.) The race-ethnic makeup of the participants was again predominantly Asian (47.0%), followed by European American (21.6%), Pacific Islander (5.2%), African American (5.2%), Latino/Latina (2.2%), Middle Eastern (0.7%), and Native American/Alaskan (0.7%). Approximately 15.7% of the participants indicated they were multiracial or of some other race-ethnicity. For Study 2, the participant-confederate sex combinations were as follows: 51 female-female, 25 female-male, 33 male-female, and 15 male-male.

The procedures for Study 2 were almost identical to those of Study 1 in the sense that participants “scored” a bogus test and relayed performance feedback to a “test taker.” However, for Study 2, the test was ostensibly of the taker’s intelligence. Moreover, a trained confederate served as the test taker/receiver for each session. Finally, in addition to receiving the laminated instruction sheet as in Study 1, scripted senders also saw their feedback message duplicated on the computer screen next to the test score. Review of the video recordings revealed that all senders in the scripted condition delivered their feedback according to the script.

As before, participants would report to the lab for an experiment on communication and test administration. After obtaining informed consent, participants were informed they would be administering a personal intelligence assessment to a second “participant” (the confederate) who, participants were told, arrived ahead of time in order to begin working on the intelligence test. Participants were then given information about the test and the scoring procedures and were left alone to review the scoring information while their partner “finished the test.”

After 3 minutes, the confederate would emerge from a nearby test-taking room and join the participant at the scoring table where the participant was already seated. Confederates would engage the participant only to say the following in a fairly subdued tone, “That was interesting. I’m a little nervous, but super curious to learn how I did.” Apart from providing polite and curt responses should the participant engage the confederate, confederates did not converse with participants.

The same basic computer scoring program from Study 1 was employed, but this time a second version of the program was developed so that the scripted message would appear on-screen when the score was revealed to the participant. As part of their lab setup, researchers would simply load either the scripted or the unscripted version of the program according to a coin flip. The scripted message was identical to that used in Study 1. As before, after the participant finished sharing the results, he or she excused the confederate to return to the test-taking room. Participants then completed a follow-up questionnaire and were debriefed, probed to discover whether they had guessed the experimental hypotheses (none indicated they had), asked not to discuss the experiment with anyone, thanked, and dismissed.

Measures

Type of test being administered

The current experimental paradigm hinges on inducing participants to believe that a test taker would want to perform well on the test and that, depending on the score, the feedback they are about to share is good news or bad news. Thus, in order to ascertain the extents to which participants might have believed each of these aspects, we included two other checks. Type of test was measured to verify the participant’s recognition that he or she was scoring and relaying results related to the test taker’s intelligence (a test on which people would want to score highly). Recognition was assessed using a single item, “For your session, which test did the test taker complete?” Participants checked one of the following: nonverbal perception, speaking ability, intelligence, or persuasion skill. All participants checked “intelligence,” thus correctly indicating the presumed nature of the test.

Test taker’s desire to know results

Consistent with previous research (Conlee & Tesser, 1973), the extent to which the participant believed the confederate really wanted to know the test result was measured to check the participant’s perception of how important the score was to the confederate. This was measured to check more intently whether the participant was buying into the importance of the test to the confederate. We developed three items to tap desire to know: “I got the sense that the test taker really wanted to know how they did on the test,” “The test taker seemed curious to learn their results on the test,” and “The test taker didn’t seem to care if they learned their score or not” (reverse scored). Participants responded using a 7-point Likert-type scale (1 = strongly disagree, 7 = strongly agree), where higher numbers correspond to a greater sense that the confederate desired to know the score (α = .87, n = 134).

Perceived news valence

Perceived news valence was assessed using the same six items from Study 1. As before, this measure demonstrated high reliability (α = .97, n = 134).

Time to response

Temporal delay before the onset of the test feedback was assessed as in Study 1.

Reluctance

For Study 2, seven items assessed sender reluctance. Two additional items, “I felt like I wanted to stall before sharing the test taker’s results with them” and “Thinking about the results I was about to share made me uneasy,” were added to those used in Study 1 on order to enhance reliability by way of internal consistency. These seven items formed a unidimensional scale that accounted for 70.83% of the variance. Cronbach’s alpha for the scale was .93 (n = 134).

Results

Induction Checks

As in Study 1, perceived news valence differed by valence category, F(2, 133) = 371.55, p < .01, η2 = .85. Further, Dunnet’s (1980) T3 post hoc comparisons revealed each mean differed from the other two, p < .01. As before, the most positive valences accompanied the highest scores, midrange valences accompanied midrange scores, and the most negative valences accompanied the lowest scores (see Table 1). Thus the induction of news valence was again effective and substantial.

A one-sample t test was performed to check the extent to which participants believed the test taker (i.e., confederate) desired to know the test results. The mean value on this variable was 5.81 (SD = 1.29), which was reliably and substantially greater than the scale midpoint of 4, t(133) = 20.77, p < .01, η² = .76. Participants indicated they believed the “test taker” was interested in learning the score.

Experimenter/Confederate Effects

Although all experimenters and confederates were trained to follow their respective standardized protocols, follow-up tests were performed to determine whether differences in the primary dependent variables were due to the specific experimenter and/or confederate. No significant experimenter or confederate effects obtained for any of news valence, time to response, or reluctance (largest F = 1.38, p > .05). Analyses proceed without regard to experimenter or confederate.

Sex Effects

No significant differences were found for news valence, F(3, 133) = 0.15, p = .93, η² < .01; or reluctance, F(3, 133) = 0.20, p = .89, η² < .01. However, sex makeup of the participant-confederate pairings did appear to influence time to response, F(3, 123) = 3.66, p < .05, η² = .08. Compared with the other three combinations, delays were slightly longer when male participants shared test scores with a female confederate. Although the proportion of variance accounted for by sex makeup was relatively small, sex makeup was added as a covariate to all analyses featuring time to response in order to control for its effects.

Tests of Hypotheses

As before, to test H1 and H2, a 3 × 2 ANOVA was performed by fully crossing three levels of news valence (superior, average, inferior) with scriptedness condition (scripted, unscripted). Planned contrasts were again used to test H3 and H4. Cases missing data were again excluded pairwise, producing slight variations in sample size. Actual Ns are reported by analysis.

H1 and H2 predicted replications of MUM effects, using time to response and reluctance as dependent variables, respectively (see Table 3). Controlling for sex makeup, the 3 × 2 ANOVA revealed the anticipated main effect for valence category on time to response, F(2, 111) = 4.16, p < .05, η² = .07. As can be viewed in Table 1, inferior scores (i.e., bad news) were delayed significantly longer than superior scores (i.e., good news). Likewise, using reluctance as the dependent variable revealed a main effect for valence category, F(2, 113) = 55.39, p < .01, η² = .49. Greater reluctance was felt when the news was bad than when the news was good (see Table 1). Thus, the Study 2 data were consistent with both H1 and 2. As in Study 1, both MUM effects replicated.

H3 (negative message–planning explanation) predicted that senders who delivered unscripted bad news would delay their responses longer than senders who delivered scripted bad news. H4 (politeness explanation) predicted that senders who delivered scripted bad news would delay their news longer than senders who delivered scripted good news. Pairwise comparisons of cell means from the 3 × 2 ANOVA using time to response revealed that unscripted bad news (n = 19) was delayed longer than scripted bad news (n = 23), t(24.26) = −2.75, p < .05, η² = .24. Further, scripted bad news (n = 23) was delayed longer than scripted good news (n = 28), t(49.00) = −2.26, p < .05, η² = .09. Thus, with regard to delay to response, the data were consistent with both H3 and H4 (see Table 4).

Senders who delivered scripted bad news were neither more nor less reluctant than senders who delivered unscripted bad news, t(43.89) = −0.53, p > .05, η² < .01. However, senders delivering scripted bad news did report feeling more reluctant than senders delivering scripted good news, t(39.00) = −8.29, p < .01, η² = .64. Thus, with respect to reluctance, these results mirror those of Study 1.

Discussion

The data from Study 2 were largely consistent with those from Study 1, with one notable exception. Whereas Study 1 failed to find a difference between unscripted bad news sharing and scripted bad news sharing (H3), Study 2 showed that scripted bad news was shared relatively sooner than unscripted bad news. It seems that scripting the news in both good and bad conditions did hasten news delivery to a small degree. However, as in Study 1, scripting the news accounted for almost zero variance in felt reluctance.

General Discussion

The purpose of this research was to explore the nature of the well-documented situation whereby one delays the interpersonal sharing of bad news. Two non–mutually exclusive explanations were tested in the context of two interaction experiments. According to the negative message–planning explanation, the sender’s “delay” is a result of increased cognitive effort involved in choosing his or her words, merely because negative messages take longer to process because of cognitive biases associated with processing negative information. The politeness explanation holds that senders delay bad news sharing strategically to mitigate perceived face threats associated with delivering the bad news. Thus, the former explanation is primarily psychological whereas the latter is primarily social.

In two separate experiments, naive test scorers shared bogus test results with a receiver, and cognitive and behavioral reactions of the test scorers were recorded. The senders’ feedback was experimentally varied to be either scripted or unscripted in a fully crossed experimental design. The data clearly and unequivocally replicated two different MUM effects (H1 and H2) for both studies. For Study 1, the data were consistent with the politeness explanation (H4) but not the verbal message–planning explanation (H3). Contrary to the prediction of H3, scripting the sender’s feedback did not significantly attenuate the MUM effect. When the news was bad (i.e., the scores were low), senders delayed the onset of their feedback longer and reported feeling more reluctant than when the news was good (i.e., the scores were high). This pattern held whether the sender’s feedback was scripted or unscripted.

Study 2 featured a design modified to raise the importance of the test (and thereby boost the strength of the news valence induction), as well as improve the time to response variable by minimizing the claim that scripted senders delayed their feedback because they needed time to locate their script. These design enhancements produced data that were largely consistent with those from Study 1. Once again, the MUM effects (H1 and H2) replicated. The reluctance data also corroborated those from Study 1 in that scriptedness condition was not found to influence the sender’s felt reluctance to share the news. The Study 2 data departed from Study 1, however, in that a main effect for scriptedness condition emerged such that scripted feedback met with shorter delays than unscripted feedback. This finding is consistent with the assumption that the delay to share news is, in part, driven by having the words to say. However, such a conclusion must be made with caution for at least two reasons. First, the effect size associated with scriptedness condition was small. Second, no interaction was found between scriptedness condition and valence category for either study (see Table 3). The appeal for scripting messages would be to attenuate some of the delay brought on by news valence. Because no interaction was found, any effects of scriptedness condition and the effects of news valence would seem to operate additively and largely independent of one another.

Thus, the data provide evidence clearly consistent with the existence of MUM effects and a politeness explanation for the delay to share bad news. The data were also more consistent with the politeness explanation than the negative message–planning explanation. The data were somewhat equivocal regarding the negative message planning–explanation in that the explanation failed to produce consistent differences. In the current research, mere negativity was neither completely falsified nor clearly supported.

Implications

Several implications follow from these results. First, this study extends MUM effect demonstrations to enlighten the process(es) underlying at least one type of MUM effect, the hesitation to share bad news. The data reported here directly inform the plausibility of two strong explanations for the delay. The results are consistent with a functional politeness explanation. This is in line with previous research. It will be recalled that Bond and Anderson (1987) observed a delay only when the sender was visible to the receiver. If planning a negative message alone begot the delay, senders should have hesitated to relay the negative feedback whether or not they were visible to the receivers because the same negative message needed to be communicated in either case. The current research results were consistent with Bond and Anderson’s findings. If increased cognitive negativity slowed verbal message production, then scripting would consistently reduce the delay. It did not.

Second, this research underscores the unique contribution of communication scholars in understanding the interpersonal delivery of bad news. Complete accounts of bad news delivery must consider both the sender as well as the target of the news. Physician-patient approaches largely focus on this process for purposes of minimizing harm to the patient. However, research shows that the bearer of the bad news often pays a considerable toll as well (e.g., Buckman, 1984; Harrison & Walling, 2010; Ptacek et al., 2001). Approaches are needed that incorporate the experiences of the sender and receiver to examine the emergent properties of the interaction. A communication perspective spotlights the components of the process that are inherent in the social interaction (e.g., the strategic communication potential of a sender’s delay of bad news).

Third, these results are valuable to theorists and practitioners from a range of specialties. Understanding the link between the delay of the feedback and the communication it represents suggests a potential strategic move for those who routinely deliver bad news (e.g., health care providers, clergy, military personnel). To the extent that the delay foreshadows the valence of the upcoming message, senders in some situations might be afforded a more reliable means of softening the impact of the news. We are already designing a study to determine if and how receivers perceive such strategic moves.

The current results also inform practitioners in at least one additional way. Although scripting specific messages might result in swifter news delivery, one thing is clear. Scripting messages seems to do nothing to alleviate the psychological reluctance senders feel when faced with delivering bad news. Two separate experiments converge on this conclusion. Thus, these findings might come as a surprise to developers of medical guidelines that routinely include rehearsing the bad news message as part of their advice. If reducing the psychological costs of sharing bad news is the goal, providing the words to say in advance does not in and of itself provide relief.

Finally, language- and message-processing specialists will also find intrigue in the delay as a functional stimulus. Communication responsibility theory (CRT; Aune, 1998; Aune, Levine, Park, Asada, & Banas, 2005) addresses the extent to which people in conversation assume responsibility for creating shared meaning. In theory, the greater the responsibility one feels, the more likely one is to be direct and the likelier it is that one will build redundancies into one’s messages (i.e., the messages will not be “lean”). Bad news sharing presents an interesting context in that a sender might delay the onset of the news as a means of “firing a warning shot” based on the negativity of the consequences for the receiver. Delaying as a means of communication is arguably less direct than sharing the bad news without delay. According to CRT, the directness of a sender’s communication is linked to the sender’s felt responsibility. Thus, the bad news delivery context is an area where the scope and boundary conditions of CRT might be refined. In fact, we are already planning a study to leverage CRT in the context of delivering bad news.

Limitations and Conclusion

Some limitations are acknowledged. Most obvious, it cannot be definitively ascertained that absolute levels of bad news or reluctance were instantiated. However, both experiments utilized a testing paradigm and metric well understood and internalized by college undergraduates. Furthermore, senders rated news valences well below the scale midpoint where appropriate, and senders’ behavior was consistent with expectations following lower or higher scores. Finally, when the social awkwardness test of Study 1 was replaced by an (ostensibly more important) intelligence test in Study 2, the news valence induction strengthened. If senders are not taking the low scores to suggest bad news, the observed patterns in the data become difficult to explain.

Another limitation concerns how time to response was operationalized. Measuring the times from the camera’s internal chronograph should be superior to using another means, such as a handheld stopwatch. However, the decision to base the stop-time on the moment the sender begins his or her first utterance to the test taker (after activating the score screen) may raise questions. For example, instead of following the initial pause with words communicating the test result, senders could simply say “Yeah,” “So,” “Okay,” or some other initial utterance, then take a few moments of silence before expressing the actual test result.⁶ Using the onset of the utterance “Yeah” as the timer stop point necessarily precludes the ability to capture what might be a conceptually meaningful pause that occurs after the sender makes such an utterance. Although it is unfortunate to miss out on these situations, capturing the delay as was done in this research ensured uniformity across cases and at least helped preserve test-retest reliability. Sadly, however, reliability by way of internal consistency could not be ascertained because the measure occurred as a single item. Single-item measures will always be less internally reliable than multiple-indicator measures, and we think this might have contributed to the attenuated effect sizes for delay relative to those associated with reluctance.

Furthermore, the generality of the findings is admittedly limited. These data were collected using university undergraduates who averaged 22 years of age. Questions are raised as to whether MUM effects and their associated phenomena are affected as people age and/or gain experience with delivering valence-sensitive information. Moreover, it is likely that the nature of the bad news in this study differs from an actual physician-patient situation in potentially important ways. The purpose of this study was to provide basic research on the nature of the temporal span whereby the sender delays the onset of the negative feedback. Because of the increased demand for internal validity, a laboratory setting served the need. Future research should determine the extent to which these findings replicate to other types of bad news as well as to real settings in which bad news is communicated.

It should be pointed out, however, that this research differs from most MUM effect demonstrations in that these samples of participants were not predominantly European American. Here, Asians and Asian Americans comprised over half of both samples. Thus, although lab experiments were performed, the convergence of these results with those derived from primarily European American samples does lend some credence to the generality of the MUM and associated effects.

Finally, these data do not inform the intentionality of the sender’s delay. It is possible that the delay is affected consciously or unconsciously. From a CRT standpoint (Aune, 1998; Aune et al., 2005), either possibility would be informative as the delay still holds the potential to activate receiver inference making (e.g., about the valence of the message) and/or to address face threats. How senders and receivers coordinate their contributions to understanding deserves further inquiry.

Understanding the process of delivering bad news from the perspective of the sender is important to those who deliver bad news on a regular basis. These data provide direct evidence that the reliable temporal delay of the onset of the bad news is more than just time needed to choose one’s words. How the receiver perceives and interprets this delay, as well as the effectiveness of the delay in mitigating face threats, are ripe areas for future research. Therefore, research that continues to explore the nature of the MUM effect is of great value.

Footnotes

Acknowledgements

The authors are grateful to Christian Gilbert, Abel Gustafson, and Kayyisa Bermudas for their assistance with data collection and to Timothy R. Levine for comments.

Authors’ Note

Jayson L. Dibble is now at the Department of Communication, Hope College, Holland, Michigan. Amy M. Wisner is now at the Department of Communication, Michigan State University. Study 1 was presented at the 2011 annual conference of the Western States Communication Association, Monterey, California.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Jayson L. Dibble (PhD communication, Michigan State University, 2008) is assistant professor in the Department of Communication at Hope College, Holland, Michigan. His research interests include interpersonal bad news delivery, alternative romantic/sexual relationship arrangements, sequential persuasion techniques, evolutionary explanations for communication behaviors, and quantitative research methods.

Amy M. Wisner received her MA in communication from the University of Hawai’i-Manoa in 2011. Currently, she is pursuing a PhD in communication at Michigan State University. Her research interests focus on interpersonal relationships and deception.

Lauren Dobbins received her MA in communicology from the University of Hawaii-Manoa in 2012. She is currently studying Persian-Farsi at the Defense Language Institute in Monterey, California. Her research interests focus on intercultural communication and networks.

Michael Cacal were MA students at the University of Hawai’i-Manoa at the time this research was conducted. They are now working in the private sector.

Emiko Taniguchi received her MA in communicology from the University of Hawai’i at Manoa in 2012. Currently, she is pursuing a PhD in communication at the University of Texas, Austin. Her research interests focus on family communication, body image, and intercultural communication.

Aili Peyton received her MA in communicology from the University of Hawaii at Manoa in 2012. Currently she is working as an editor for Anchorage Daily News, Anchorage, AK.

Lisa van Raalte received her MA in communicology from the University of Hawaii at Manoa in 2012. She now works as an educational research associate for the Research Corporation of the University of Hawaii. She plans to pursue further doctoral work.

Andra Kubulins earned her MA from the University of Hawaii in communicology, and a BA in communication from Michigan State University. She is currently living in Boston, MA.

References

Aune

R. K.

(1998, July). A theory of attribution of responsibility for creating understanding. Paper presented at the annual meeting of the International Communication Association, Jerusalem, Israel.

Aune

R. K.

Levine

T. R.

Park

H. S.

Asada

K. J. K.

Banas

J. A.

(2005). Tests of a theory of communicative responsibility. Journal of Language and Social Psychology, 24, 358-381. doi:10.1177/0261927X05281425

Baumeister

R. F.

Bratslavsky

Finkenauer

Vohs

K. D.

(2001). Bad is stronger than good. Review of General Psychology, 5, 323-370.

Berscheid

(1983). Emotion. In Kelly

H. H.

(Ed.), Close relationships (pp. 110-168). New York, NY: Freeman.

Bisel

R. S.

Kelley

K. M.

Ploeger

N. A.

Messersmith

(2011). Workers’ moral mum effect: On facework and unethical behavior in the workplace. Communication Studies, 62, 153-170. doi:10.1080/10510974.2010.551314

Bisel

R. S.

Messersmith

A. S.

Kelley

K. M.

(2012). Supervisor-subordinate communication: Hierarchical mum effect meets organizational learning. Journal of Business Communication, 49, 128-147.

Bond

C. F.

Jr. Anderson

E. L.

(1987). The reluctance to transmit bad news: Private discomfort or public display? Journal of Experimental Social Psychology, 23, 176-187.

Brown

Levinson

S. C.

(1987). Politeness: Some universals in language usage. Cambridge, England: Cambridge University Press.

Buckman

(1984). Breaking bad news: Why is it still so difficult? British Medical Journal, 288, 1597-1599.

10.

Conlee

M. C.

Tesser

(1973). The effects of recipient desire to hear on news transmission. Sociometry, 36, 588-599.

11.

Dibble

J. L.

Levine

T. R.

(2010). Breaking good and bad news: Direction of the MUM effect and senders’ cognitive representations of news valence. Communication Research, 37, 703-722.

12.

Dibble

J. L.

Levine

T. R.

(2012). Sharing and bad news with friends and stranger: Reasons for and communication behaviors associated with the MUM effect. Manuscript submitted for publication.

13.

Diener

Emmons

R. A.

(1985). The independence of positive and negative affect. Journal of Personality and Social Psychology, 47, 1105-1117.

14.

Dillard

J. P.

Segrin

Harden

J. M.

(1989). Primary and secondary goals in the production of interpersonal influence messages. Communication Monographs, 56, 19-38.

15.

Dunnet

C. W.

(1980). Pairwise multiple comparisons in the unequal variance case. Journal of the American Statistical Association, 75, 796-800.

16.

Eggly

Penner

Albrecht

Cline

R. J. W.

Foster

Naughton

Ruckdeschel

J. C.

(2006). Discussing bad news in the outpatient oncology clinic: Rethinking current communication guidelines. Journal of Clinical Oncology, 24, 716-719.

17.

Fallowfield

Jenkins

(2004). Communicating sad, bad, and difficult news in medicine. The Lancet, 363, 312-319.

18.

Fisher

C. D.

(1979). Transmission of positive and negative feedback to subordinates: A laboratory investigation. Journal of Applied Psychology, 64, 533-540.

19.

Fiske

S. T.

(1980). Attention and weight in person perception: The impact of negative and extreme behavior. Journal of Personality and Social Psychology, 38, 889-906.

20.

Goffman

(1967). Interaction ritual. New York, NY: Pantheon Books.

21.

Harrison

M. E.

Walling

(2010). What do we know about giving bad news? A review. Clinical Pediatrics, 49, 619-626. doi:10.1177/0009922810361380

22.

Hatfield

Cacioppo

J. T.

Rapson

R. L.

(1992). Primitive emotional contagion. In Clark

M. S.

(Ed.), Review of personality and social psychology (Vol. 14, pp. 151-177). Beverly Hills, CA: Sage.

23.

Ito

T. A.

Larsen

J. T.

Smith

N. K.

Cacioppo

J. T.

(1998). Negative information weighs more heavily on the brain: The negativity bias in evaluative categorizations. Journal of Personality and Social Psychology, 75, 887-900.

24.

Mehrabian

(1966). Silent messages. Belmont, CA: Wadsworth.

25.

Peeters

Czapinski

(1990). Positive-negative asymmetry in evaluations: The distinction between affective and informational negativity effects. European Review of Social Psychology, 1, 455-474.

26.

Ploeger

N. A.

Kelley

K. M.

Bisel

R. S.

(2011). The hierarchical mum effect: A new investigation of organizational ethics. Southern Communication Journal, 76, 465-481.

27.

Ptacek

J. T.

Ptacek

J. J.

Ellison

N. M.

(2001). “I’m sorry to tell you . . .” Physician’s reports of breaking bad news. Journal of Behavioral Medicine, 24, 205-217.

28.

Rosen

Tesser

(1970). On reluctance to communicate undesirable information: The MUM effect. Sociometry, 33, 253-263.

29.

Rozin

Royzman

E. B.

(2001). Negativity bias, negativity dominance, and contagion. Personality and Social Psychology Review, 5, 296-320.

30.

Sussman

S. W.

Sproull

(1999). Straight talk: Delivering bad news through electronic communication. Information Systems Research, 10, 150-166.

31.

Taylor

S. E.

(1991). Asymmetrical effects of positive and negative events: The mobilization-minimization hypothesis. Psychological Bulletin, 110, 67-85.

32.

Tesser

Rosen

(1975). The reluctance to transmit bad news. In Berkowitz

(Ed.), Advances in experimental social psychology (Vol. 8, pp. 193-232). New York, NY: Academic Press.

33.

Tesser

Rosen

Tesser

(1971). On the reluctance to communicate undesirable messages (the MUM effect): A field study. Psychological Reports, 29, 651-654.

34.

Uysal

Oner-Ozkan

(2007). A self-presentational approach to transmission of good and bad news. Social Behavior and Personality, 35, 63-78.

35.

Weenig

M. W. H.

Wilke

H. A. M.

ter Mors

(2011). Personal outcomes and moral responsibility as motives for news transmission: The impact of fate similarity, fate uncertainty, and relationship closeness. Communication Research. Advance online publication. doi:10.1177/0093650211402195

36.

Wilson

S. R.

Aleman

C. G.

Leatham

G. B.

(1998). Identity implications of influence goals: A revised analysis of face-threatening acts and application to seeking compliance with same-sex friends. Human Communication Research, 25, 64-96.

37.

Wilson

S. R.

Kim

M. S.

Meischke

(1991). Evaluating Brown and Levinson’s politeness theory: A revised analysis of directives and face. Research on Language and Social Interaction, 25, 215-252.

38.

Yariv

(2006). “Mum effect”: Principals’ reluctance to submit negative feedback. Journal of Managerial Psychology, 21, 533-546.