Introduction

Deviant workplace behaviors present both organizational and societal problems. Behaviors such as theft, sabotage, misrepresentation of performance, and excessive absenteeism can all reduce the effectiveness and efficiency of organizations (Greenberg 2002; Lau et al. 2003; Murphy 1993). From a societal perspective, poor organizational performance can adversely impact employment levels and the stability of national and global economies (Cochran 1964; Noe and Rebello 1994). The cost of workplace deviance is difficult to quantify, and while estimates vary considerably, most studies suggest that the price of deviance is substantial. A report by the Association of Certified Fraud Examiners (2012) suggested that employee dishonesty, examined in our first study, may cost employers as much as $3.5 trillion USD globally. Some industries such as medical care, which serves as the context for our second study, are particularly vulnerable. Deviation from recognized standards of practice routinely results in financial penalties that, in the U.S., totaled about $373 million in 2014 (DHHS 2014).

A number of perceptual factors, such as perceived injustice and inequity, have been linked to deviant workplace behaviors (e.g., Ambrose and Schminke 1999; Aquino et al. 2001; Greenberg 1990; Lau et al. 2003; Mars 1973, 1974). To date, the perceptual and emotional processes that underlie these relationships have received relatively limited direct empirical attention (e.g., Aquino et al. 2001; Douglas and Martinko 2001; Greenberg 1990). Using two studies, we attempt to examine and clarify the nature of the relationships between attributions, moral emotions, and behaviors by examining the predictive power of Weiner’s (1985a) attributional model in three ways. First, we utilize affective events theory (AET: Weiss and Cropanzano 1996) as a macrostructure for linking attributional perceptions to moral emotions and deviant behaviors. In so doing, we also respond to Weiss and Beal’s (2005) call for research aimed at identifying the cognitive processes that underlie the AET framework. Second, a comprehensive study design is used to test the influence of each possible combination of high and low levels of three attributional dimensions on moral emotions and behavioral responses. This effort is in line with recent calls to clarify the links between attributions, emotions, and workplace behaviors (Dasborough et al. 2011; Martinko et al. 2011). Third, the examination is conducted on two highly distinct samples (i.e., undergraduate business students and practicing physicians) so that the influence of different experiential and contextual factors on the predictive power of attributions and moral emotions can be compared.

Moral Emotions and the Justification of Deviant Behavior

Organizational deviance refers to deliberate actions that are intended to harm others, violate rules or norms, and reduce organizational performance (Geddes and Callister 2007; Neuman and Baron 1997; Lau et al. 2003; Robinson and Bennett 1997). Examples include theft (Greenberg 1990, 2002; Mars 1973, 1974), withholding effort (Kemper 1966), sabotage (e.g., Ambrose et al. 2002), and misrepresentation of performance (Schweitzer et al. 2004).

We posit that the justification of such behavior involves an employee reaching the conclusion that deviance is somehow warranted given his or her specific circumstances. This does not imply that employees believe the behaviors to be permissible or organizationally condoned. Rather, justification allows individuals to believe that their behavior is reasonable and defensible in their specific situation. In the present study, we examine the role of moral emotions in the cognitive process that enables this justification to occur.

Numerous researchers have investigated factors that promote retaliatory workplace deviance (e.g., DeMore et al. 1988; Douglas and Martinko 2001; Greenberg 1990; Kemper 1966; Mars 1973, 1974). Although the justification process is not the explicit focus of these studies, some of the conclusions and findings provide insight into how individuals might justify deviant behaviors. Several authors explicitly or implicitly invoke equity and justice theories to explain why employees might decide to steal, withhold effort, or overstate their performance. Studies by Mars (1973, 1974) and Greenberg (1990, 2002), for instance, have observed that employees often justify workplace theft when they feel they are being underpaid. Neuman and Baron (1997) referred to this form of deviance as instrumental in that its goal is to right a perceived wrong rather that to cause harm, although harm may be a secondary motivation (see also: Ambrose et al. 2002; Jones and Skarlicki 2005; Skarlicki and Folger 1997). Conversely, expressive deviance is driven by the primary goal of causing harm to perceived wrongdoers (Kemper 1966; Neuman and Baron 1997).

A common theme among these theoretical arguments and empirical findings is that deviant acts are often driven by negative emotional reactions to perceived wrongdoing. In the case of expressive deviance, negative affect can override rational cognitive processes, whereas instrumental deviance involves logical retaliation that is often triggered by a negative affective experience (Martinko et al. 2005). To this point, Weiss et al. (1999) argued that emotions play a particularly important but understudied role in shaping deviant behaviors. The subset of discrete affective states known as moral emotions appears to be particularly relevant, as suggested by Haidt’s (2003) characterization of these emotions as those that are triggered by moral violations (e.g., pay cuts that are perceived to be unfair; Greenberg 1990). In support of this notion, several studies indicate that the moral emotion of anger is often associated with workplace deviance (Barclay et al. 2005; Douglas and Martinko 2001). We build on this knowledge with a comprehensive investigation of the relationship between attributions, moral emotions, and deviant behavior.

Attributional and Emotional Predictors of Deviant Behavior

Our hypothesized relationships, depicted in Fig. 1, are based on the premise that attribution-driven moral emotions can influence individuals’ ability to justify deviant behavioral responses to negative workplace events. Based on AET (Weiss and Cropanzano 1996) and Weiner’s (1985a) framework, it is predicted that attributions linked to the moral emotions of anger and shame will facilitate the justification of deviant behaviors, whereas those linked to guilt will attenuate justification.

Fig. 1
figure 1

Hypothesized relationships

Full size image

AET was developed as a “macrostructure” for understanding the impact of workplace emotions and, as Weiss and Beal (2005, p. 4) noted, “Intended for researchers to give greater attention to events, their interpretation, their structure [and] their informational value.” The framework suggests that positive and negative affective workplace events (e.g., a favorable or unfavorable performance evaluation) trigger emotional states that influence employees’ attitudes and behaviors. Importantly, AET recognizes situation-specific factors (e.g., organizational climate, supervisor behaviors) that can help shape these reactions as we discuss in our contextual factors section below.

We suggest that attribution theory, and Weiner’s (1985a) model in particular, is well suited to explain a number of the processes that underlie AET. Consistent with AET’s focus on affective trigger events, attribution theory is based on the notion that individuals seek to understand the causes of significant events in their lives, particularly when they are important, negative, and/or unexpected (Heider 1958; Weiner 1985b). This causal reasoning process has been linked to a wide array of affective and behavioral outcomes (see Harvey et al. 2014 for a meta-analytic review).

Again consistent with AET, Weiner’s (1985a) framework posits that the affective event-behavior relationship is mediated by emotions. More specifically, Weiner (1985a) argued that attributions for important events initiate emotional reactions that drive behavioral responses. The type of emotional reaction that develops is influenced, in part, by the dimensionality of the attribution. The locus of causality dimension describes the extent to which an event is attributed to causes internal or external to the observer (e.g., skill vs. help from a coworker) and is particularly relevant to the valence of an emotion. The stability dimension refers to the perceived permanence or variability of a cause (e.g., intelligence vs. effort) and has affective implications through its influence on expectations regarding future outcomes. A third dimension, controllability, describes the extent to which the cause of an event is perceived to be deliberate or haphazard and can impact emotions both directly, similar to the locus dimension, and indirectly, similar to the stability dimension (Harvey et al. 2014).

Consideration of these dimensions allows us to develop arguments concerning the manner in which attributions for affective events can help explain both the emotional and behavioral reactions predicted by AET. For example, negative trigger events that are attributed to externally controllable factors (e.g., deliberate unfair treatment by a supervisor) are predicted to promote feelings of anger, whereas personally controllable attributions (e.g., insufficient effort) are thought to provoke feelings of guilt. Weiner (1985a) also predicted that attributing negative events to personally uncontrollable factors (e.g., lack of intelligence) would promote shame.

The moral emotions resulting from these three attributional dimensions are associated with different behavioral responses. Anger, for instance, has been linked to acts of aggression, retaliation, and sabotage (Kemper 1966; Skarlicki and Folger 1997; Weiss et al. 1999). Shame has been linked to psychological withdrawal, whereas guilt has been linked to constructive behaviors such as increased effort and pro-social behavior (Bracht and Regner 2013; Weiner 2004). Building on these findings, our basic thesis is that AET and the attribution–emotion–behavior framework can, at least in part, explain how and when employees are able to justify deviant workplace behaviors.

Hypothesis Development

Several studies suggest that employees may find deviant behaviors justifiable if they feel their employing organization has engaged in moral violations against them (DeMore et al. 1988; Douglas and Martinko 2001; Greenberg 1990, 2002; Mars 1973, 1974). From an attributional perspective, five aspects of these studies appear to be relevant. First, each case of deviance studied was preceded by a negative affective trigger event capable of initiating a causal search. Second, in each study, the most proximal cause of the events from the subjects’ perspective was the employing organization, an external and relatively stable factor. Third, the studies indicate that subjects felt their employers were deliberately exploiting them, denoting attributions of external control. Fourth, the authors of these studies typically made at least a passing reference to the fact that employees harbored negative emotions toward their organizations as a result of the trigger event. Finally, the target of subjects’ deviant behaviors was generally the same external and stable factor—the organization—thought to have caused the trigger event.

The consistencies across these studies suggest that there is a relationship between attributing negative workplace events to stable, organizationally controllable factors and the justification of deviant behavior. However, because these studies did not specifically test this hypothesis, it must be considered speculative. As such, the present study is designed to explicitly test the influence of attributions and moral emotions on the justification of deviant behavior.

Anger-Producing Attributions

The subjects in the Greenberg (1990, 2002), DeMore et al. (1988), and Mars (1973, 1974) studies were all confronted with negative affective events. AET suggests that an emotional reaction will follow such events and Weiner’s (1985a) logic helps us predict what form that reaction will take. The participants in each study appeared to cite stable and externally controllable factors (i.e., organizations and their leaders) when justifying their deviant behavior; Weiner’s (1985a) logic indicates that these types of attributions promote feelings of anger toward the external entity (see also, Weiner 1987). The stability of the attributions can exacerbate the anger response because the perceiver believes that he or she cannot do anything to remove the stable causal factor (e.g., Campbell and Martinko 1998; Douglas and Martinko 2001).

Attribution theory can also aid in predicting the behavioral consequences of affective trigger events and the associated emotional reactions. Driven by external attributions, anger is a particularly strong moral emotion and is thought to have more influence on behaviors than less intense emotions (e.g., Fischer et al. 2004; Geddes and Callister 2007; Geddes and Stickney 2011). If the perceived causes of anger are also stable in nature, AET suggests that the recurrent individual emotional events can influence immediate affect-driven behavioral responses (e.g., expressive deviance) as well as instrumental behaviors driven by negative, anger-driven work attitudes that develop over time.

Although the behavioral implications of felt or expressed anger are not always negative (Lindebaum and Fielden 2011), angry individuals are generally more likely to act on their emotions than those who are mildly annoyed or frustrated (e.g., Douglas et al. 2008; Douglas and Martinko 2001; Geddes and Lindebaum 2014). This is consistent with AET’s notion of affect-driven emotion and with research on motivated reasoning. The latter suggests that more affectively arousing trigger events can promote bias-prone, affect-dominant “hot” cognitions that are more likely to provoke impulsive and potentially reckless behavior than more elaborative “cold” cognitions (e.g., Kunda 1990; Redlawsk et al. 2010). In the case of perceived workplace wrongdoings and moral violations, we predict that anger will promote the former, facilitating the justification of deviant responses.

Hypothesis 1a

The attribution of undesirable affective workplace events to stable and organizationally controllable factors is positively associated with feelings of anger.

Hypothesis 1b

Via this attributional process, anger mediates the positive relationship between undesirable affective workplace events and the justification of deviant behavior.

Shame-Producing Attributions

We predict that the attribution of negative affective workplace events to stable and personally uncontrollable factors, such as low intelligence, will also be associated with deviant behavior. Weiner’s (1985a) model indicates that this type of attribution is likely to cause shame by making individuals feel that they lack the power to remedy a personal shortcoming that caused the event.

Shame can be triggered by many of the same affective events that promote anger, such as poor performance reviews and being passed over for promotions. Tracy and Robins (2006) noted that internal, uncontrollable, and stable attributions for such events can promote isolated incidents of affect (shame)-driven behaviors, whereas a chronic tendency to form such attributions can manifest itself as negative self-related attitudes and a pattern of shame-induced behavior.

In either case, we expect that the deviance resulting from shame is likely be of the instrumental as opposed to expressive variety. Potentially representing a “colder,” more elaborative thought process, shame is predicted to motivate behaviors intended not to harm others but to prevent them from learning about their failures or their responsibility for these failures. That is, employees may perceive that the benefits of hiding their shame (e.g., by concealing mistakes) outweigh the costs (e.g., the risk of getting caught and potentially harming innocent parties; Tripp and Bies 1997).

Based on these arguments, we predict that negative affective workplace events promote shame when they are attributed to stable and personally uncontrollable factors. Shame, in turn, is predicted to facilitate the justification of deviant behaviors, such as misrepresenting performance or covering up past mistakes.

Hypothesis 2a

The attribution of undesirable affective workplace events to stable and personally uncontrollable factors is positively associated with feelings of shame.

Hypothesis 2b

Via this attributional process, shame mediates the positive relationship between undesirable affective workplace events and the justification of deviant behavior.

Guilt-Producing Attributions

Prior research and theory suggest that desirable responses to negative affective workplace events are likely to occur when individuals attribute them to factors that are personally controllable and unstable, such as insufficient effort (Abramson et al. 1978; Tracy and Robins 2006). These attributions are often associated with “adaptive guilt” because they cause individuals to take responsibility for the affective event and to become aware of steps that could have been taken to prevent it (Tracy and Robins 2006; Weiner 1985a; Weiss et al. 1999).

Guilt can therefore facilitate desirable responses because the personally controllable attributions that promote it can be addressed through constructive means such as increased effort (Perrewé and Zellars 1999; Weiss et al. 1999). A recent study by Ent and Baumeister (2015) also suggests that guilt is associated with a desire to avoid harm to others. In another study, Cohen (2010) found that guilt was associated with heightened disapproval of unethical negotiation practices, suggesting an inability to justify such tactics. Taken together, this research suggests that guilt may promote judgment-driven, rather than impulsive affect-driven responses to negative workplace events.

Because these attributions provide no external target for deviant behaviors, and because they suggest remedies through ethical means such as increased effort, unstable and personally controllable attributions for negative events are not predicted to promote the justification of deviant behavior. Thus, although guilt is a negative and unpleasant emotion, we predict that it is negatively associated or uncorrelated with deviant behavior.

Hypothesis 3a

The attribution of undesirable affective workplace events to unstable and personally controllable factors is positively associated with feelings of guilt.

Hypothesis 3b

Via this attributional process, guilt mediates the negative relationship between undesirable affective workplace events and the justification of deviant behavior.

Personally controllable causes are typically unstable in nature but may be stable in some cases. A person’s education level, for instance, is relatively stable but still personally controllable. Thus, if an individual believes she is passed over for a promotion because she does not have an advanced degree, she is making a stable yet personally controllable attribution. As with unstable and personally controllable attributions, we argue that feelings of guilt are likely to result. We again predict that guilt will help individuals believe that constructive behavioral responses (e.g., increasing one’s education level) are possible, making a deviant response difficult to justify.

Hypothesis 4a

The attribution of undesirable affective workplace events to stable and personally controllable factors is positively associated with feelings of guilt.

Hypothesis 4b

Via this attributional process, guilt mediates the negative relationship between affective workplace events and the justification of deviant behavior.

Additional Attributional Conditions

The preceding hypotheses represent arguments we are able to formulate using existing research and theory. In the interest of developing and testing a complete research design and maximizing the explanatory power of the study, we also tested the impact of the other possible combinations of the three attributional dimensions on emotions and behavioral justification. More specifically, we looked at the impact of attributions for negative affective workplace events caused by stable or unstable and organizationally uncontrollable factors (e.g., laws or economic fluctuations), unstable and personally uncontrollable factors (e.g., losing out on a sale because of a temporary illness), and organizationally controllable unstable factors (e.g., temporary wage freezes) on emotions and deviant responses.

Contextual Factors

The attribution–emotion–behavior framework is thought to reflect relatively universal human cognitive patterns (Weiner 1985a). We recognize, however, that workplace characteristics and other contextual factors can influence each of the relationships depicted in Fig. 1. In their development of AET, Weiss and Cropanzano (1996) argued that aspects of the workplace environment could influence the initial appraisal of affective trigger events as well as the emotional and behavioral responses to the events. They noted that numerous situation-specific variables are relevant in this regard, including many that are difficult to control for (e.g., ambient noise and temperature levels).

As with Weiss and Cropanzano (1996; see also Weiss and Beal 2005), our primary focus is on the cognitive process surrounding affective events rather than situational factors. We felt, however, that testing our hypotheses in a single context would create a risk of producing results that might not generalize to other populations and contexts. We therefore utilized two sample populations that differed significantly in terms of several key variables that have been linked to moral emotions and ethical behavior. These included age, education levels, experience, socialization, and occupational/organizational culture (Borkowski and Ugras 1998; Hershfield et al. 2012).

The first sample consisted of undergraduate business students, representing lower levels of age, education, and experience relative to the second sample, which consisted of practicing physicians. In terms of socialization, the physicians had all experienced numerous years of acculturation in their medical training and in their current practices, whereas undergraduate students typically have lower levels and more varied forms of organizational socialization. The organizational and occupational culture into which physicians are socialized is particularly unique from an ethics perspective, given the importance of the Hippocratic oath and the professional, moral, and legal consequences of inappropriate behavior.

These individual and contextual factors can reasonably be expected to influence the extent to which subjects allow their behavior to be driven by moral emotions. Although we made no formal hypotheses regarding differences between these two samples, our general expectation is that negative emotional responses to attributions may be less pronounced and less strongly associated with deviance in the physician sample.

Methods

Procedure

A 2 × 2 × 2 scenario-based study was designed to test our hypotheses. Despite their history of use in studies of ethics and deviance, scenario designs have generated some controversy. As Cavanagh and Fritzsche (1985) observed, properly designed scenarios can create a relatable context in which researchers can examine behavioral tendencies that are not easily observed in real-world situations. The deviant behaviors studied here fit these criteria given their socially undesirable and potentially career-threatening nature. Webber (1992), however, observed that a lack of methodological and analytical rigor in a number of published studies had damaged the reputation of scenario designs. Over 20 years later, we observe that similar concerns remain salient in the field of organizational studies. Webber explained that these flaws were not inherent in the nature of scenarios, however, and argued that their potential benefits could be realized when scenarios are designed and analyzed properly.

Following Webber’s (1992, pp. 153–156) guidelines, we developed two sets of eight scenarios through a series of pilot studies. Each scenario depicted one of the possible attributional conditions consisting of high or low levels of externality (i.e., organizational or personal cause), stability, and controllability. In each scenario, the subjects were confronted with a negative affective trigger event that might realistically occur in their organizations. In the student sample, subjects were told that they received a poor grade on an assignment; in the physician sample, subjects were told that they failed to order a required test for a patient.

Participants were randomly assigned to one of the eight attributional conditions. In the first part of each scenario, subjects were informed of the negative affective event and its cause (summarized in Table 1). A manipulation check was accomplished by asking subjects to rate the cause of the event as internal or external, stable or unstable, and controllable or uncontrollable. Subjects also indicated their emotional reaction at this stage. In the second part of Study 1, student subjects were informed that their hypothetical professor lost their grades and were asked to tell the professor what their grades were. Subjects were told that they could tell the professor that they received a higher grade than they actually earned without fear of getting caught. In Study 2, the physician subjects were told that they had the opportunity to alter their paperwork to make it appear that the test was ordered on time and avoid being sanctioned for negligence.

Table 1 Summary of attributional manipulations
Full size table

Given the involvement of patient health, the seriousness of the deviance was unavoidably higher in the physician sample. To at least partially offset the impact of this on physicians’ responses, participants in each condition were informed that “Although the delay in ordering the test did not place the patient in a life-threatening situation, it is considered an improper standard of care.” Physician participants were also told that, in this specific instance, there was no risk that the alteration of the paperwork would be detected.

Participants

Study 1

Four hundred and forty-seven students (71.6 % response rate) enrolled in multiple sections of a management course at a US university over the course of a year participated in Study 1. The sample consisted of 266 females (59.7 %) and 181 males (40.5 %), with a mean age of 21.6 years. Subjects were instructed to report to a computer lab where they registered for nominal course credit. They were then randomly assigned to a computer terminal hosting one of the eight scenario-based exercises, each representing a different experimental condition.

Study 2

The Study 2 sample consisted of 121 physicians (17.2 % response rate) practicing in the southeastern United States. Each subject was employed by the same healthcare provider, which consented to the use of its members for this study. The sample consisted of 34 females (28.1 %) and 87 males (71.9 %), with a mean age of 55.1 years. To recruit subjects, the president of the health care facilities where the physicians were employed sent an email to each physician. The message detailed the purpose of the study and provided a link to one of the eight anonymous web-based surveys. The distribution list was split into eight sub-lists by arranging the name of each potential respondent alphabetically and breaking the list into eight groups of equal size. Each group was assigned to one of the eight experimental conditions.

Measures and Manipulations

Causal Attribution

Attributions for negative trigger events were manipulated using the hypothetical scenarios described above and summarized in Table 1. An example of one causal manipulation, the stable and organizationally controllable manipulation, involved telling Study 1 subjects that their professor was very unclear about the requirements for the assignment for which they received a poor grade and refused to clarify these requirements. For the same manipulation in Study 2, subjects were told that they failed to order a required test for their patient because management chronically understaffs their department.

Attributions made in response to the hypothetical events were measured for manipulation check purposes using items adapted from the Organizational Attribution Style Questionnaire (Kent and Martinko 1995). A sample item (used to measure locus of causality) was “To what extent do you believe that the poor grade you received (Study 1)/failure to order the test (Study 2) was caused by something about you versus your professor (Study 1)/employer (Study 2)?” Responses were scored on seven-point scales (Locus of causality: 1, “Totally due to me,” 7, “Totally due to my professor/employer”; Stability: 1, “Never present,” 7, “Always present”; Control: 1, “No control,” 7, “Total control”).

Emotional Response

Anger and guilt were measured using items from the PANAS-X scale developed by Watson and Clark (1994). This scale asks subjects to rate their levels of different emotional reactions to the outcomes described in the scenarios on a 5-point scale (1, “Very slightly or not at all,” 5, “Very Much”). Each emotion was measured using four items. The items used to measure anger (Study 1 α = .86, Study 2 α = .94) were “Angry,” “Disgusted,” “Scornful,” and “Hostile.” Items used to measure guilt (Study 1 α = .84, Study 2 α = .91) were “Guilty,” “Angry at self,” “Blameworthy,” and “Disgusted with self.” Because the PANAS scale does not specifically measure shame, items similar to the others in the scale were developed using a thesaurus to generate synonyms for the term. The resulting items (Study 1 α = .86, Study 2 α = .93) were “Ashamed,” “Embarrassed,” “Humiliated,” and “Regretful.”

Justification of Deviant Behavior

Subjects’ ability to justify the deviant behaviors presented in the scenarios was assessed by rating the behavior as justifiable or unjustifiable using three items (Study 1 α = .80, Study 2 α = .90) modified from Reidenbach and Robbin’s (1990) scale (see also, Spicer et al. 2004). Responses were recorded on a 7-point scale (1, “Definitely not justifiable,” 7, “Definitely justifiable”). A sample item included “Saying that you earned a higher grade on the project is a just response to this situation.”

Control Variables

Age and gender were controlled in the study, based on research indicating that both can influence emotional reactions and ethical tendencies (Ambrose and Schminke 1999; Brody and Hall 2000; Erikson 1963). Trait-level negative affectivity (NA) was also controlled to account for the impact of predispositions toward negative emotional responses. It was measured using Watson and Clark’s (1994) scale prior to presenting subjects with the scenarios (Study 1 α = .89, Study 2 α = .91). This scale asked respondents to rate the frequency with which they experience the range of positive and negative emotions contained in the measure “in general” in order to assess the trait form of NA.

Results

Manipulation Check

Locus of causality scores (external conditions: Study 1 mean = 4.37, Study 2 mean = 3.94; internal conditions: Study 1 mean = 2.07, Study 2 mean = 2.50) differed significantly between the internal and external groups (Study 1: F(1,446) = 378.89, p < .001, Study 2: F(1,120) = 23.40, p < .001). The same was true for stability scores (stable conditions: Study 1 mean = 4.30, Study 2 mean = 3.93; unstable conditions: Study 1 mean = 3.37, Study 2 mean = 3.34), although the p value for Study 2 was slightly above .05 (Study 1: F(1,446) = 45.96, p < .001, Study 2: F(1,120) = 3.33, p < .07) and controllability scores (controllable conditions: Study 1 mean = 5.08, Study 2 mean = 4.80; uncontrollable conditions: Study 1 mean = 4.58, Study 2 mean = 4.08) as well as (Study 1: F(1,446) = 12.06, p < .05, Study 2: F(1,120) = 5.15, p < .05).

Mean Comparisons

Means, standard deviations, and correlations are shown in Tables 2 and 3. Between-condition mean differences are shown in Table 4. Consistent with our expectations, levels of reported anger were highest among subjects in the stable and externally uncontrollable attributional case in Study 1 although the mean differences for this emotion were not significant in Study 2. The justification of a deviant response was also highest in this condition in the student sample and second highest in the physician sample. Levels of reported shame were relatively high among Study 1 participants in the condition thought to promote shame (stable, personally uncontrollable attributions) but were slightly higher in the two conditions associated with guilt. Mean differences for this emotion were again insignificant in Study 2. Guilt levels were highest in the two attributional conditions thought to promote this emotion in both studies.

Table 2 Means, standard deviation, and correlations among study variables (Study 1)
Full size table
Table 3 Means, standard deviation, and correlations among study variables (Study 2)
Full size table
Table 4 Mean differences
Full size table

Hypothesis tests were conducted using Hayes and Preacher (2014) mediation procedure with sequential group coding. This technique accommodates categorical independent variables and multiple mediators, allowing us to test each of the hypothesized relationships simultaneously. Gender, age, and negative affectivity were entered as control variables in order to minimize spurious variance in the emotion or justification variables associated with these factors. In the following sections, we summarize the direct effects of attributions on each moral emotion (see also Table 5) and their mediated effects on the justification of deviance.

Table 5 Attribution–emotion relationships
Full size table

Hypothesis 1: Attributions, Anger, and Justification

Hypothesis 1a predicted that the attribution of negative affective trigger events to stable and organizationally controllable causes would promote anger. Study 1 results supported this hypothesis, indicating a significant positive association between this attributional manipulation and anger (β = .44, p < .01) as shown in Fig. 2. A similar relationship was observed in Study 2 although it did not reach significance in this sample (β = .45, p = .09).

Fig. 2
figure 2

Results of hypothesis tests (Bold figures sample 1, Unbolded figures sample 2; IE indirect effect; **p < .01, *p < .05)

Full size image

Hypothesis 1b predicted that anger mediates the relationship between these attributions and the justification of deviant behavior. In Study 1, a significant association between anger and justification was observed (β = .54, p < .01) and the attributional manipulation demonstrated a significant indirect effect on justification through anger (effect: .24, 95 % CI: .07,.44), supporting the mediation hypothesis. The hypothesis was not supported in the physician sample given the lack of a direct attributional effect and a non-significant relationship between anger and justification in this group.

Hypothesis 2: Attributions, Shame, and Justification

Hypothesis 2a predicted that the attribution of negative affective events to stable and internally uncontrollable factors would promote shame. A positive association between this attributional manipulation and shame was observed in Study 1 (β = .67, p < .01) and Study 2 (β = .79, p < .05) supporting the hypothesis. In Study 1, however, a significant association with shame was also observed between unstable and internally controllable attributions (β = .44, p < .05). As noted above, this attributional combination has been conceptually linked to guilt. We discuss this finding in greater detail below.

Hypothesis 2b predicted that shame would mediate the attribution–justification relationship. A mediated effect was observed in Study 1 (effect: −.17, 95 % CI −.35, −.03) in which shame was associated with the justification of deviance. Unexpectedly, however, the impact of shame on justification was negative (β = −.25, p < .01). A significant mediation effect was not observed in Study 2.

Hypotheses 3 and 4: Attributions, Guilt, and Justification

Hypotheses 3a and 4a predicted that unstable or stable, personally controllable attributions for negative affective events would be associated with feelings of guilt. Study 1 results supported Hypothesis 3a, suggesting that unstable, personally controllable attributions for negative events would promote feelings of guilt (β = .67, p < .01). In Study 2, however, this attributional manipulation was more strongly associated with anger (β = .75, p < .05), suggesting a possible contextual effect on the relationship, as we discuss below. Stable and personally controllable attributions were not associated with guilt in either study, contrary to Hypothesis 4a.

Hypotheses 3b and 4b predicted that guilt would mediate the impact of these attributions on the justification of deviant behavior. Guilt was not significantly associated with justification in either sample, however, suggesting that it neither promoted nor inhibited deviance.

Exploratory Relationships

Results suggested that several of the exploratory attributional conditions were significantly associated with emotional outcomes. Unstable and organizationally uncontrollable attributions for negative trigger events showed a strong negative relationship with guilt in both studies and shame in Study 1, as indicated in Table 5. The stable and externally uncontrollable attribution condition was associated with both shame and guilt in Study 1 and no emotional outcomes in Study 2, as shown in Table 5. The Study 1 findings were unexpected given the external nature of the attributions as discussed below. The final two attributional conditions showed a good deal of divergence between the two samples as indicated in Table 5, again suggesting the possible presence of a contextual effect, which we discuss in the next section. Of particular interest was the apparent salience of unstable, organizationally controllable attributions in the student sample. As the table indicates, this group demonstrated significant reactions with all three emotional responses—heightened anger levels and reduced shame and guilt levels. Conversely, this attributional manipulation had no emotional impact in the physician sample.

Discussion

The studies summarized above investigated the extent to which causal attributions for, and emotional reactions to, negative workplace events facilitate the justification of deviant behavior. Although our results indicated general support for the notion that emotion-driven attributions can impact the justification of deviance, it is clear that sample characteristics and context had a significant impact on the strength and nature of these relationships.

This discussion begins with a summary of results grouped by emotional outcome. Between-sample differences are then discussed, with an emphasis on situational and cognitive factors that might account for divergent results.

Anger

We anticipated that stable, externally controllable attributions for negative affective trigger events would promote anger, which, in turn, would facilitate the justification of deviant behavior. This attribution–emotion–behavior relationship was observed in the Study 1 student sample but not in the Study 2 physician sample. While we anticipated that effect sizes might be muted in the latter sample, we were somewhat surprised to see the relationship completely disappear. While the smaller sample size and associated reduction in statistical power in Study 2 may have accounted for some of this attenuation, we expect that contextual factors and population characteristics were also involved, as we discuss below.

Shame

Stable and personally uncontrollable attributions for negative affective events were predicted to promote feelings of shame and did so in both samples. This attribution-driven emotion was only associated with justification in the student sample, however, and the effect was the opposite of what was predicted. We hypothesized that shame would facilitate the justification of deviance but instead found a negative association. While this finding is logical in the sense that the cause of the negative events was internal to the subjects, it contradicts our logic that the desire to hide the cause of shame would promote the justification of a deviant response. A similarly unexpected finding was also reported by Cohen (2010), who observed that shame-prone individuals were more likely than others to approve of unethical negotiation tactics (i.e., lying and false promises). These findings might be explained by Tripp and Bies’ (1997) observation that deviant behaviors are often seen as undesirable when innocent bystanders are likely to be affected. Because shame is associated with internal causes for negative events, retaliatory responses that impact individuals who are not responsible for the undesirable trigger events may therefore be difficult to justify.

Also unexpected was the significant impact of unstable, personally controllable attributions (thought to promote guilt) on feelings of shame in Study 1. Guilt and shame are both internally directed negative emotions so some degree of overlap is to be expected (Tangney et al. 1992). This is illustrated in the similar shame and guilt effects observed in most of the exploratory attributional conditions (see Table 5). Three of these four conditions involved internal attributions for negative trigger events, making a self-directed emotion such as guilt or shame more likely than emotions such as anger, which are often externally focused. The fact that these findings were not observed in Study 2, however, suggests that the student subjects may have been more prone to report a blend of the two emotions.

Guilt

Guilt was associated with unstable and personally controllable attributions in the student sample, as predicted, but not in the physician sample. A number of exploratory attributional manipulations were also negatively associated with guilt, as shown in Table 5. Not surprisingly, each of these effects was observed in conditions where the negative workplace event was attributable to external factors. Guilt was not associated with the justification of deviance in either sample. This contradicted our prediction that the emotion would reduce justification levels but supported the notion that guilt does not promote the justification of deviance.

Population Effects

The between-sample differences we observed were generally in line with our informal prediction of muted effect sizes in the physician sample as compared to the student sample, although the contrast was more pronounced than expected. In this section, we discuss contextual and individual factors that might account for the differences in emotional reactions and deviance between the two samples.

Socialization

In comparing undergraduate students to practicing physicians, differences in socialization levels and experience appear important. A study by Rayburn and Osman (2004) suggested that physicians are subject to a number of unique expectations, including the expectation that they harbor a forgiving and merciful personality. To conform to this expectation, physicians may be socialized to refrain from retaliating against others who have caused negative events to occur.

AET posits that contextual factors such as these can encourage reasoned, judgment-driven behavioral responses as opposed to more spontaneous, affect-driven behaviors (Weiss and Cropanzano 1996). Attribution theory has also evolved to recognize that situational factors can exacerbate or attenuate the degree to which affective reactions shape the rationality and spontaneity of attributional and behavioral responses (Douglas et al. 2008). Like AET, Douglas et al. (2008) acknowledged that pre-existing workplace norms and attitudes could alter the strength and nature of emotional and attributional responses to affective trigger events. In the case of medical workplace norms, at least one ethicist (Thompson 2004) has argued that angry outbursts by physicians are a violation of medical ethics because patients’ needs, as opposed to personal feelings, are expected to receive priority.

Thus, through socialization members of the physician sample may have developed cognitive schema that led them toward judgment versus affect-driven behaviors in response to the negative trigger events. Consistent with this reasoning, our results suggest that physicians were less likely to form or act upon negative moral emotions than members of the student sample. This may help explain why moral emotions were far more strongly associated with attributions and justification in the student sample.

Socialization mechanisms that influence the severity of emotional, and ultimately behavioral, reactions to undesirable workplace events may take numerous forms. For instance, physicians might be selected and evaluated based, in part, on their demonstrated ability to maintain low and inelastic levels of negative emotions in difficult situations. Further, their training and informal socialization is likely to reinforce the importance of emotional stability. Students, or members of other professions where the experience and display of negative emotions is tolerated to a higher degree (e.g., construction; Lindebaum and Fielden 2011), may report a wider range of discrete emotions. In these emotionally tolerant situations, emotions may play a larger role in the shaping of behaviors.

Emotional Intelligence

Building on the potential role of restrained emotional reactions, we also note the possibility of between-sample variance in emotional intelligence levels. Salovey et al. (2004) conceptualization of emotional intelligence involves, among other aspects, the ability to understand and manage emotions. The latter ability may be particularly important regarding affective reactions to negative trigger events. It may, for example, allow employees to hold the strength of anger responses below the level at which impulsive, affect-driven acts of deviance become more likely. Research generally posits that the development of emotional intelligence is influenced by patterns of interactions and external influences (e.g., cultural) that occur over the course of individuals’ lives (e.g., Zeidner et al. 2003). This logic suggests that emotional intelligence levels may have been higher in the physician group, a possibility that could help explain the muted negative emotions and associated deviance levels in this population.

Ethicality

A third between-sample consideration is the potential for different levels of ethicality among the participants of the two studies. Although the notion that physicians could be more ethical on average than college students might appear overly broad, there is some evidence suggesting that this may the case. It might, for instance, be argued that the physician participants possessed a higher degree of moral development given their age and education levels (Colby et al. 1983; Kohlberg 1969; Kohlberg and Candee 1984; Paradice and Dejoie 1991; Rest 1986).

Moral development is argued to promote ethical behavior by raising individuals’ awareness of the negative impact of unethical behavior and is thought to increase with age and education level (Kohlberg 1969). The mean age of the physician subjects was approximately 34 years older than that of the student subjects and the physicians were also more educated. Research also indicates that physicians demonstrate higher moral development scores than members of most other professions (Coleman and Wilkins 2004). As such, moral development may account for the weaker relationships between attributions, moral emotions and deviance in Study 2.

Moral disengagement

Another potential source of between-sample variation is differences in levels of moral disengagement. Moral disengagement is the ability to suspend behavioral self-regulation mechanisms and allows individuals to engage in unethical behaviors without experiencing high levels of guilt (Bandura 1991; Bandura et al. 2001). Bandura (1991) identified a number of moral disengagement methods including exonerative comparison in which individuals attempt to diminish the perceived severity of their own harmful behaviors by comparing them to more egregious behaviors of others. It is possible that a number of the student subjects had witnessed behaviors among their peers that were more egregious than the performance misrepresentation depicted in the scenarios, making the hypothetical behavior seem more justifiable by comparison. Conversely, the relatively high levels of moral development thought to exist among physicians (Coleman and Wilkins 2004) might reduce the ability of these subjects to look to others’ behaviors as a means of moral disengagement.

Additional methods of moral disengagement involve denying one’s responsibility for the harm resulting from a deviant behavior (Bandura 1991). This technique is particularly relevant to this study because it deals with issues of perceived causality. The fact that external attributions for negative events were only associated with heightened justification levels in the student sample suggests that the physician subjects may have been less able to overlook their own responsibility for their behaviors. This appears to be a logical conclusion, given the amount of trust and personal responsibility given to physicians by patients and administrators. Thus, the Study 1 participants in the external cause conditions might have seen deviant responses, and their consequences, as ultimately being the responsibility of the external causal entity, whereas the physicians focused on the consequences of their behaviors rather than the external stimuli that preceded them.

The possibility that student subjects engaged in moral disengagement more easily than the physicians also has implications for the emotional component of our hypotheses. Bandura et al.’s (2001) study indicated that moral disengagement is associated with affective rumination, a process that causes individuals to develop feelings of anger toward perceived wrongdoers that can trigger retaliatory behavior. From an AET perspective, this rumination may elevate the power of the affective (anger) response in predicting the behavioral outcome. Thus, moral disengagement might help explain the stronger propensity toward both anger and the justification of deviance in the student sample.

Implications for Preventing Workplace Deviance

This study suggests that, in at least some situations, understanding individuals’ emotional and attributional responses to affective trigger events may be important for preventing deviant behavior. As noted in the introduction, workplace deviance is associated with significant organizational and societal costs. An understanding of the cognitive processes and contextual factors associated with deviance can therefore be valuable if it allows employers to attenuate deviance and the associated expenses.

Our findings, particularly in the student sample, suggest that steps taken to promote accurate causal assessments of negative affective workplace events such as poor evaluations or missed deadlines might reduce the likelihood of impulsive, affect-driven deviance. Communicating the causes of negative events, even if they seem obvious to managers, may help toward this goal. If the true cause of such an event is one that might facilitate the justification of a deviant response (e.g., a relatively stable cause), additional steps such as helping employees diffuse negative emotions and devise constructive and ethical responses may be advisable.

Evidence from our physician sample speaks, at least indirectly, to the impact of socialization on both emotional and behavioral reactions to negative affective workplace events. While AET observes the power of the workplace environment to impact these reactions, the wholesale implementation of an organizational culture that promotes measured emotional reactions and judgment-driven behaviors is likely beyond the capacity of most managers. Nevertheless, efforts at encouraging such tendencies through selection, evaluation, and training, however, may gradually promote a culture that expects and rewards thoughtful responses to negative trigger events rather than the impulsive, affect-driven reactions that are more likely to take deviant forms (Douglas et al. 2008).

Our findings also underscore the need for employees to believe that non-deviant options exist to remedy difficult workplace situations. Among otherwise rule-abiding employees, deviance may be seen as a behavior of last resort by those who feel they have been placed in an unjust situation. While the effectiveness of any given technique can vary considerably between organizations, various mechanisms for promoting employee “voice” (e.g., ombudsman roles, feedback meetings) may help employees explore non-deviant remedies to negative affective events. Conversely, when opportunities for employee voice are perceived to be absent, the likelihood of deviant responses to negative workplace events appears to rise (Vries et al. 2012). In these situations, stable and uncontrollable attributions are more likely, potentially giving employees a sense of desperation where deviant responses appear justifiable (Harvey et al. 2009).

Contributions, Limitations, and Future Research Suggestions

This study provides some insight into the cognitive process through which individuals justify deviant behaviors. The study also extends past research on attributions and emotions by investigating the full range of attributions composed of three established attributional dimensions, and the extent to which these attributions can promote the justification of deviant behavior. It also adds to past attributional research by comprehensively testing Weiner’s (1985a) attribution–emotion–behavior framework. This model has served as a basis for a good deal of attributional research although some of the predicted relationships have not been investigated in previous empirical studies.

There are also several limitations to this research that could be addressed in future research. First, the scenario design places subjects into hypothetical situations in which true reactions can only be estimated. Constructive replications of this study in both field and experimental settings would be valuable. We also focused on the justification, as opposed to the actual performance, of deviant behavior. Thus, rather than using hypothetical scenarios in future research, it might be helpful to have subjects recall instances where they engaged in actual deviant behavior and recall the attributions and emotions associated with the behaviors.

Second, the low response rate for the physician sample suggests that some degree of response bias may have been present. Although a number of factors likely limited this response rate (e.g., lack of participation incentives as compared to the course credit offered to the student sample and less time available to participate in the study), it is possible that a reluctance to participate in a workplace deviance study deterred some potential participants. The fact that the invitation to participate was issued by organizational leadership may have exacerbated this reluctance. While it is difficult to speculate on how this type of response bias might have influenced our results, the possibility of such an influence must be acknowledged.

Third, the study considered only three attributional dimensions (locus, stability, and controllability). These three dimensions form the basis of Weiner’s (1985a, b) framework, which we sought to empirically examine here, but are not the only ones relevant to affective reactions. The intentionality dimension in particular has been linked to emotional responses (Dasborough and Ashkanasy 2002). An investigation of this dimension in the context of negative affective events and deviance would be a natural extension of the present study. Finally, negative moral emotions other than those indicated by Weiner’s (1985a) framework might be relevant to the justification of deviance. Future research could include a wider array of emotions to test this possibility.

The results of this study and the post hoc explanations for the unanticipated results described above suggest a number of additional avenues for future research. In terms of possible socialization effects, a clear extension of this research is to study other distinct populations that vary in the type and level of socialization their members’ experience. Emotional intelligence levels could also be measured to determine, if as speculated above, they account for some degree variance in emotional and behavioral reactions between and within sample populations. The same is true for the other individual difference variables discussed previously, such as ethicality/moral development and moral disengagement propensities.

Conclusions

This study built on the premise that we could begin to understand how employees justified deviant workplace behavior by examining the attributions and moral emotions associated with deviance. We found that, at least in some instances, aspects of this general assumption appear valid, but that context and population-specific characteristics likely play an important role.

The contrast between the physician sample and the student sample strikes us as particularly interesting. Our tentative explanations for many of these differences are based on the likelihood that the training, ethicality, and experience of the physicians may have enabled less deviant, judgment-driven reactions to negative workplace events. The speculative nature of these arguments underscores the need for a greater understanding of how affective and attributional processes interact with situational factors, not only in the context of deviant behavior but in the more general context of organizational behaviors.

We believe that this research makes two primary contributions. First, the results suggest that we can begin to develop a better understanding of how people justify deviant behaviors by examining attributional and emotional processes. Second, it begins to delineate the conditions under which individuals might engage in judgment-driven versus emotion-driven processing of negative workplace events. Further research in this area will be necessary if we are to more fully understand how and when people are able to justify and engage in deviant behaviors. This same work would also contribute to our more general knowledge of organizational behaviors. It is our hope that the findings presented here will serve as a basis for future research toward this goal.