Fairness is widely recognized by organizational scholars as an important antecedent of workplace attitudes and behaviors including job satisfaction, organizational commitment, trust, organizational citizenship behaviors, counterproductive work behaviors, absenteeism, theft, and job performance (Cohen-Charash & Spector, Reference Cohen-Charash and Spector2001). For this reason, identifying the social and cognitive mechanisms that underlie fairness perceptions represents an important area of organizational research. Organizational justice researchers have generally assumed that fairness is assessed by examining the extent to which treatment is consistent with societal norms and values (e.g., Colquitt & Greenberg, Reference Colquitt, Greenberg and Greenberg2003). Therefore, it is not surprising that a large literature has emerged demonstrating that internalized cultural norms and values influence people’s definitions of what types of events represent justice violations. For example, cross-cultural differences in uncertainty avoidance (Otto, Baumert, & Bobocel, Reference Otto, Baumert and Bobocel2011) have been implicated as important determinants of what people view as unfair.
Despite the importance attributed to differences in cross-cultural values, the extant literature has not considered how related cross-linguistic norms might influence fairness judgments. Cross-linguistic differences have only recently emerged in the social sciences as a viable mechanism for understanding how people perceive and interact with the world. According to Evans and Levison (Reference Evans and Levinson2009), this research was largely stinted by a bias among social scientists that the more than 5000 languages in the world are English-like with different verbs and nouns. This bias appears to stem from the almost ubiquitous ability to directly translate between languages. However, languages, like cultural values, are learned through repeated exposure to norms. As a result, the linguistic syntax used to encode events in one language often differs from how the same event would be described in a second language. These cross-linguistic norms, rather than the language per se, have been the subject of recent research examining how language affects cognition. For example, in the Australian aboriginal language Guungu Yimithirr, object positions are always described using an absolute, or cardinal, frame of reference (e.g., “The tree is north of the house”), rather than the relative frame of reference (e.g., “The tree is in front of the house”) common in most languages (Levinson, Reference Levinson2003). Although this might seem like a minor difference in how people describe object positions, this research has found that the ubiquitous use of cardinal directions endows native Guungu Yimithirr speakers with an uncanny sense of direction, even in unfamiliar locations. Similar effects of cross-linguistic norms on cognitions have been implicated in a variety of contexts including how people perceive shapes, colors, and time (e.g., Imai & Gentner, Reference Imai and Gentner1997).
Of particular interest to the current article is research examining cross-linguistic differences in the use of agentive (e.g., “she broke the vase”) versus non-agentive syntax (e.g., “the vase broke”). In contrast to English, Spanish linguistic norms dictate that non-agentive syntax be used to distinguish accidental from intentional events (Choi, Reference Choi and Gathercole2009). Building on prior research, Fausey and Boroditsky (Reference Fausey and Boroditsky2010; Reference Fausey and Boroditsky2011) proposed and empirically demonstrated that the use of non-agentive syntax (1) makes it more difficult for Spanish speakers to identify the parties responsible for mistakes, and (2) attenuates accountability judgments. Fausey and Boroditsky argued that these effects occur because the responsible party is not explicitly referenced in the non-agentive syntax used by Spanish speakers, which makes the responsible party less salient when encoding and recalling the mistake.
In this article, we suggest that these findings have three potential implications for understanding how language affects cognitions and behavioral intentions in reaction to justice violations. First, several theories of organizational justice propose that fairness is an inherently social construct that requires the identification of a responsible party (e.g., Folger & Cropanzano, Reference Folger, Cropanzano, Greenberg and Cropanzano2001). Folger and Cropanzano go so far as to state that, “If no one is to blame, there is no social injustice” (p. 1). In turn, non-agentive syntax might represent a rhetoric device for attenuating perceptions that an event is unfair and subsequent intentions to hold the responsible party accountable. Footnote 1 Second, cross-linguistic norms dictate that Spanish speakers encode accidents using non-agentive syntax, whereas English speakers encode accidents using agentive syntax. Thus, when justice violations are perceived to occur accidentally, Spanish speakers might be less likely to perceive the event as unfair and also they might be less likely to hold the responsible party accountable.
Third, in addition to the effects of language on event fairness perceptions, it is possible that language also influences how entity fairness information is used in reaction to accidental justice violations. Choi (Reference Choi2008) proposed and showed that when the entity responsible for a discrete violation of justice is considered to be generally fair, people give the entity the benefit of the doubt and are less likely to hold the responsible party accountable than when the entity is considered to be generally unfair. However, if non-agentive syntax makes the responsible party less salient, it might also make characteristics about the entity, including entity fairness information, less salient. In turn, qualifying Choi’s findings, when accidental justice violations are encoded using Spanish, entity fairness information might not affect behavioral intentions to hold the responsible party accountable.
Therefore, we extend the organizational justice literature by examining how cross-linguistic norms (i.e., the use of agentive versus non-agentive syntax) influence cognitions and behavioral intentions in reaction to accidental justice violations. Specifically, we examine (1) how non-agentive versus agentive syntax influences event fairness perceptions and behavioral reactions to justice violations (Experiment 1), (2) how the language (i.e., English versus Spanish) used to encode accidental justice violations influences fairness perceptions and behavioral intentions (Experiment 2), and (3) how the language used to encode the event moderates the influence of entity fairness perceptions on behavioral intentions (Experiment 3). Experiment 3 also investigates our claim that these effects result from differences in cross-linguistic norms beyond other factors. We do this by randomly assigning bilingual participants to a Spanish or English condition and controlling for native language and cross-cultural norms investigated in previous research. As far as we know, this is the first research to examine how language influences workplace relevant criteria.
Agentive syntax and fairness
In addition to suggesting that fairness perceptions are formed by assessing how consistent treatment is with cultural norms, several organizational justice scholars have suggested that fairness is an inherently social construct (e.g., Lind & Tyler, Reference Lind and Tyler1988). For example, van Prooijen, van den Bos, and Wilke (Reference van Prooijen, van den Bos and Wilke2005) stated, “justice is a social phenomenon, because experiences of (in) justice are products of people’s interactions with others in social settings” (p. 664). Building on the inherently social nature of fairness perceptions, several dominant theories of organizational justice have proposed that identifying a responsible social entity is fundamental to fairness evaluations (e.g., Folger & Cropanzano, Reference Folger, Cropanzano, Greenberg and Cropanzano2001; Mikula, Reference Mikula2003). These theories posit that fairness assessments are primarily used to determine whether a social entity should be held accountable for experienced treatment, and empirical research has generally supported these predictions. For example, Nicklin, Greenbaum, McNall, Folger, and Williams (Reference Nicklin, Greenbaum, Mcnall, Folger and Williams2011) found that fairness judgments are related to perceptions that a viable social entity was responsible for the event. Similarly, Naquin and Kurtzberg (Reference Naquin and Kurtzberg2004) demonstrated that when a non-social entity (i.e., the technology itself) was portrayed as responsible for a technological failure, perceptions that the treatment was unfair were reduced relative to descriptions that emphasized the responsibility of a social entity (i.e., members of the IT department). In addition, related research has suggested that explanations can disperse the blame attributed to viable social entities, thereby attenuating perceptions that the experienced treatment was unfair (Colquitt & Chertkoff, Reference Colquitt and Chertkoff2002).
Despite the importance placed on the identification of a responsible party and the influence of explanations on fairness perceptions and subsequent accountability behaviors, the extant organizational justice literature has not explored how the language used to describe and encode events influences reactions to justice violations. Therefore, we begin by examining the potential influence of descriptions that use agentive versus non-agentive syntax on fairness perceptions and subsequent behavioral intentions.
We should note that politicians seem to be well aware that certain syntactic devices can be used to displace blame. For example, after the BP oil spill, President Obama used an unusually large number of passive sentences when speaking about himself or his Administration, whereas references to BP were primarily agentive (Parker, Reference Parker2010). Ronald Reagan went one step further removing the agent entirely in his famous quotation about the Iran-Contra scandal, “Mistakes were made.” In addition, empirical research has demonstrated that syntactic variations in descriptions of events can reduce the blame attributed to the responsible party by making the responsible party less salient when encoding and recalling events. For example, Kasof and Lee (Reference Kasof and Lee1993) asked participants to read one of two sentences: (1) Ted hit Bob; (2) Bob was hit by Ted. These authors found that moving the responsible agent to the end of a sentence (i.e., changing the sentence from active to passive) decreased the salience of Ted and reduced perceptions that Ted was responsible for the event.
Fausey and Boroditsky (Reference Fausey and Boroditsky2010; Reference Fausey and Boroditsky2011) extended this research by investigating whether similar effects occur when the agent is removed entirely from the event descriptions (i.e., non-agentive syntax). Fausey and Boroditsky (Reference Fausey and Boroditsky2010) asked participants to read a description of the Justin Timberlake/Janet Jackson “wardrobe malfunction” that occurred during the Super Bowl XXXVIII halftime show. The results of this experiment revealed that participants who read agentive descriptions of event (e.g., Justin tore the bodice) viewed Justin as more responsible and indicated that Justin should be fined 53% more than participants who read non-agentive descriptions of the same event (e.g., the bodice tore). Consistent with Kasof and Lee (Reference Kasof and Lee1993), these authors suggested that the language used to encode the event influenced the salience of the responsible party. If fairness perceptions require the identification of a responsible social entity, then explanations of justice violations that use non-agentive syntax should be perceived to be less unfair than the same violation described using agentive syntax. Moreover, if fairness assessments are primarily used to determine whether the responsible party should be held accountable, fairness perceptions will mediate the relationship between agentive versus non-agentive syntax and intentions to hold the responsible party accountable for justice violations.
Hypothesis 1: Non-agentive descriptions of justice violations will reduce intentions to hold the responsible party accountable relative to agentive descriptions of the same violation.
Hypothesis 2: Fairness perceptions will mediate the influence of non-agentive versus agentive descriptions of justice violations on intentions to hold the responsible party accountable.
Cross-linguistic differences in reactions to unfair treatment
In addition to the potential influence of non-agentive syntax as a rhetoric device that can attenuate negative reactions to justice violations, variation in cross-linguistic norms that dictate the use of agentive versus non-agentive syntax when encoding mistakes might also influence fairness perceptions and subsequent behavioral intentions. Although not all incidents of unfair treatment are accidental, unfair treatment frequently results from accidental rather than intentional motives. For example, strategic mistakes can lead to layoffs, mistakes in scheduling can lead to unfair distributions of tasks, and mistakes concerning who is responsible for a product’s performance can lead to an unfair distribution of bonuses. In addition, researchers have suggested that the fairness of an event is often ambiguous (e.g., Folger & Cropanzano, Reference Folger, Cropanzano, Greenberg and Cropanzano2001). Therefore, it is possible that even intentional incidents of unfairness are often perceived to be accidental. Thus, identifying how language influences reactions to accidental justice violations seems to represent an important extension of the organizational justice literature.
Linguistic analyses have demonstrated that certain languages such as Spanish and Japanese utilize non-agentive language to distinguish accidental from intentional actions (Choi, Reference Choi and Gathercole2009). For example, in a series of studies, Fausey and Boroditsky (Reference Fausey and Boroditsky2010; Reference Buhrmester, Kwang and Gosling2011) found that native English speakers who viewed videos of people making mistakes (e.g., accidently breaking a glass) were more likely to describe the event using agentive language (e.g., she broke the glass); whereas native Spanish and Japanese speakers were more likely to use non-agentive descriptions (e.g., el vaso se rompió). Further, Fausey and Boroditsky (Reference Fausey and Boroditsky2011) found that English speakers were more likely to hold the responsible party accountable in a subsequent task. In addition, Fausey and Boroditsky provided evidence that these effects resulted from the relative salience of the responsible party. Specifically, compared to English speakers, Spanish speakers had difficulty identifying the party responsible for a mistake even a few minutes after witnessing the mishap. Consistent with Fausey and Boroditsky’s findings, we predict that when native English speakers witness accidental justice violations they will be more likely to encode the violation using agentive syntax compared to native Spanish speakers who will be more likely to encode the same violation using non-agentive syntax. In turn, the responsible party should be less salient to Spanish speakers causing the event to seem less unfair compared to native English speakers. In turn, English speakers should express greater intentions to hold the responsible party accountable, relative to native Spanish speakers. Specifically, we hypothesize the following:
Hypothesis 3: Compared to English speakers, Spanish speakers will be more likely to encode accidental justice violations using non-agentive language, which will attenuate perceptions that the event was unfair and also decrease behavioral intentions to hold the responsible party accountable.
Language and entity fairness
In addition to the direct effects language (i.e., Spanish versus English) might have on event fairness perceptions, we suggest that the use of non-agentive syntax when accidental justice violations are encoded in Spanish also moderates the effect of entity fairness information on behavioral intentions to hold the responsible party accountable. Cropanzano, Byrne, Bobocel, and Rupp (Reference Cropanzano, Byrne, Bobocel and Rupp2001) were the first to suggest that there are two distinct paradigms in the organizational justice literature. The most frequently employed paradigm assesses fairness in reaction to a particular justice violation (Bies & Shapiro, Reference Bies and Shapiro1987). Cropanzano et al. (Reference Cropanzano, Byrne, Bobocel and Rupp2001) referred to this paradigm as event fairness.
On the other hand, the social entity paradigm asks participants to respond to the general fairness of entities such as an organization, a supervisor, or a coworker. In contrast to the event-based perceptions assessed using scales, such as Colquitt’s (Reference Colquitt2001) measures of organizational justice (e.g., “Does your [outcome] reflect the effort you have put into your work?”), the social entity paradigm asks participants to respond more generally and to rely on characteristics of a social entity rather than a specific event. For example, Niehoff and Moorman (Reference Niehoff and Moorman1993, p. 541) asked participants to indicate the extent to which “All job decisions are applied consistently across all affected employees.” Although these two paradigms have historically been investigated independently, Choi (Reference Choi2008) provided the first empirical research integrating the two paradigms.
Specifically, Choi (Reference Choi2008) argued that the influence of unfair events (i.e., justice violations) on subsequent perceptions of trust and intentions to engage in organizational citizenship behaviors (OCBs) are influenced by entity fairness perceptions. Choi suggested that when an entity is generally considered to be fair, people give the social entity the benefit of the doubt when they violate justice norms. On the other hand, if the social entity is generally considered to be unfair, negative reactions to justice violations are amplified. In order to assess the developed model, Choi (Reference Choi2008) surveyed supervisor-employee pairs. Consistent with Choi’s model, the results of the survey revealed that entity fairness perceptions moderated the influence of event fairness on subsequent trust and OCBs directed toward the entity.
Although Choi (Reference Choi2008) provided valuable insights into the relationship between event and entity fairness, we argue that the cross-linguistic differences in the use of agentive versus non-agentive syntax influences the use of entity fairness information when reacting to accidental justice violations. As mentioned earlier, the salience of the responsible party when encoding and recalling mishaps is presumed to be responsible for cross-linguistic differences in recall and accountability judgments (Fausey & Boroditsky, Reference Fausey and Boroditsky2010; Reference Fausey and Boroditsky2011). Therefore, we predict that when accidental justice violations are encoded using Spanish, the relatively low salience of the responsible party created by encoding the event with non-agentive syntax will also make characteristics about the party, including entity fairness information, less salient. Thus, qualifying Choi’s (Reference Choi2008) findings, we suggest that when Spanish speakers experience an accidental justice violation they will be less influenced by entity fairness when judging trust in the entity and intentions to engage in OCBs that will help the responsible party.
Hypothesis 4: The relationship between entity fairness and behaviors that hold the responsible party accountable for accidental justice violations (i.e., OCBs) will be moderated by the language used to encode the event.
Overview of the studies
We conducted three experiments to test each of our hypotheses. (See Figure 1 for a visual description of the hypothesized models and the associated experiments.) In the first experiment, English-speaking participants were asked to read a scenario in which agentive and non-agentive syntax was used. In the scenario, a mistake was made and an employee received a negative outcome; participants were then asked to report their fairness perceptions. The rationale of this first experiment is to show that the differences in blame and fairness are explained by how participants encode what they read or see. People who encode the actions agentively are more likely to ascribe intentions to these actions (and are more likely to perceive events as less unfair).
Figure 1. Hypotheses and Studies.
Our second experiment used a sample of native English- and Spanish-speakers. Participants were asked to watch a video involving a scenario with similar characteristics to the scenario used in Experiment 1. Our rationale was that because English-speaking individuals are more likely to speak, write and encode events agentively, they are more likely to blame others, and perceive events as unfair, than Spanish-speakers. We tested here Hypothesis 3, which posits that Spanish-speakers would use non-agentive language to encode the accidental justice violation, and that this encoding would mitigate perceptions of unfairness and subsequent behavioral intentions to hold the responsible party accountable (i.e., OCB intentions). The third and final experiment involves a similar video, but we used bilingual participants randomly assigned to interact with the experiment in Spanish or English. In this experiment we tested Hypothesis 4, which predicts that language moderates the relationship between entity fairness and the degree to which the entity is perceived to be blameworthy. In Experiment 3 we also sought to disentangle the effects of language from potential effects of cultural norms and native language.
EXPERIMENT 1
Method
Participants
A sample of 118 students from a large university in the Southwest participated in this experiment (average age = 21.6, percentage of males = 47%). Participants were monolingual English speaking U.S. citizens recruited from a large university in the Southwest. All participants indicated that over 80% of their current daily language use is English.
Design and procedures
Participants were randomly assigned to either the Agentive or Non-Agentive condition. Specifically, participants were asked to read a scenario describing the same accidental justice violation using either agentive or non-agentive syntax. After reading the scenario, participants were asked to complete measures of fairness perceptions, and behavioral intentions to hold the responsible party accountable.
Scenarios
The following scenarios were used to communicate an accidental justice violation. These scenarios were adapted from prior research by Nicklin and Williams (Reference Nicklin and Williams2009). Participants were asked to place themselves in the shoes of the student when reading the scenario. Underlining and italics are used to highlight differences in the scenarios for the reader of this article, but were not included in the versions read by participants.
Participants in the Agentive condition read: “At the beginning of the semester you received academic advising and later found out that the advisor did not provide you with the correct information. The advisor made a mistake and registered you for the wrong course. You are a senior and will have to stay an extra semester to complete the correct course in the Fall.”
Participants in the Non-Agentive condition read: “At the beginning of the semester you received academic advising and later found out that you were not provided with the correct information. A mistake was made and you were registered for the wrong course. You are a senior and will have to stay an extra semester to complete the correct course in the Fall.”
Measures
Fairness perceptions
Fairness perceptions were measured using a three-item scale taken directly from Nicklin and Williams (Reference Nicklin and Williams2009; α = .83), which was adapted from Ambrose and Schminke’s (Reference Ambrose and Schminke2009) measure of overall justice. An example item was “Overall, the treatment I received was fair.” (1 = strongly disagree; 7 = strongly agree).
Behavioral intentions
The one-item measure of behavioral intentions was also taken directly from Nicklin and Williams (Reference Nicklin and Williams2009). Specifically, participants were asked to indicate, “How likely is it that you would write a formal complaint letter to be put in the advisor’s permanent personnel file?” (1 = very unlikely; 7 = very likely).
Results and Discussion
In order to investigate the effects of agentive descriptions on fairness perceptions and behavioral intentions to hold the responsible party accountable, we first performed a MANOVA. Fairness perceptions and behavioral intentions were entered as the dependent variables and agentive versus non-agentive language was entered as the independent variable. The results revealed that compared to participants who read the non-agentive scenario, participants who read the agentive scenario were significantly more likely to view the mistake as unfair F(1, 116) = 7.42, p < .01, η2 = .06, and to suggest that they would send a letter reprimanding the academic advisor F(1, 116) = 8.05, p < .07, η2 = .07. Table 1 presents the independent t-tests and means across condition for each of the dependent variables. These results support Hypothesis 1, which predicted that the use of non-agentive syntax when describing justice violations would decrease intentions to hold the responsible party accountable when compared to agentive descriptions of the same violation.
Table 1. Means, Standard Deviations, and t-tests for Experiment 1
Note: **represents significance at p < .01.
Next, we directly tested Hypothesis 2, which predicted that the use of agentive versus non-agentive syntax would influence intentions to hold the responsible party accountable via fairness perceptions. We used bootstrapping procedures described by Preacher and Hayes (Reference Preacher and Hayes2007). The bootstrapping method has been considered as one of the best (most powerful) tests of mediation (Kenny, Reference Kenny2016). In support of Hypothesis 2, the results revealed that the 95% confidence interval surrounding the indirect effect did not include zero (b = –.06; S.E. = .04; 95% CI: –.18, –.01). The results of this experiment suggest that agentive language can influence reactions to unfair mistakes even when the participants are native English speakers, and suggest that non-agentive syntax is a powerful rhetoric device for attenuating negative reactions to justice violations. However, a limitation of this experiment is that in the non-agentive condition the subject of the action was not mentioned. Thus, it is somewhat expected that blame and fairness perceptions were so different across conditions. In Experiment 2, we address this issue by using a scenario in which we used a video; therefore, the encoding was made by participants, not suggested by researchers.
EXPERIMENT 2
In this experiment we sought to examine the idea that even if people are presented with virtually the same information, those who speak and think in Spanish (as opposed to English) encode the information differently; thus, they perceive the situation differently in terms of fairness.
Method
Participants
Experiment 2 investigated Hypothesis 3. Thirty-six native English speakers (average age = 29.6 years; 53% males) and thirty-four native Spanish speakers (average age = 28.1; 79% males) participated in the experiment. Consistent with prior linguistics research (Fausey & Boroditsky, Reference Fausey and Boroditsky2011) participants were monolinguals recruited via Amazon’s Mechanical Turk service (https://www.mturk.com) Footnote 2 . A mechanical turk advertisement was set up to solicit native English and Spanish speakers separately using the appropriate language in each advertisement. The scope of the advertisement was limited to individuals whose IP addresses indicated that they were currently within in the United States. The limitation to participants currently in the U.S. was intended to provide a stronger test of the assertion that language rather than cross-cultural values are responsible for the hypothesized effects compared to using Spanish and English speakers living in different nations (Experiment 3 provides a more explicit test of this assertion). All participants indicated that they were native English or Spanish speakers consistent with the targeted demographic and all participants indicated that they used their native language over 80% of the time on a daily basis. Participants read and listened to instructions in either English or Spanish consistent with their native language. Instructions were developed by the first author, translated into Spanish by the second author - who is a native Spanish speaker but is bilingual - and verified by an independent bilingual (i.e., a native English speaker).
Procedure
All participants were instructed that they would watch a video about a hypothetical company. Participants viewed a manager writing out bonus checks on the basis of employee performance reports. The manager places the wrong check on top of the wrong performance report resulting in an objectively unfair distribution of bonuses. Specifically, the manager places a check with the name I. Jones on top of Julia Jones’ performance report and a check with the name J. Jones on top of Iris Jones’ performance report. Although Iris Jones received a 2.5/5 on her performance report, Iris Jones receives the larger of the two bonuses ($2000 rather than $200) because the manager accidently places her check on top of Julia Jones’ performance report, who received a 4.5/5. Conversely, although Julia Jones received a 4.5/5 on her performance report the manager accidently pays Julia Jones the smaller of the two bonuses ($200 rather than $2000) because he accidently places her check on top of Iris Jones’ performance report, who had received a 2.5/5. The English version of the video can be viewed at (http://lofvideos.participantpool.com/?ver=Study1English and the Spanish version can be viewed at http://lofvideos.participantpool.com/?ver=Study1Spanish).
We should note that there are no differences in the description of the video: both the English version and the Spanish version used the same script and syntax. In fact, at no point in the videos (Spanish and English) it is explicitly mentioned that there is mistake; or if there is anyone to blame. The narrator reads: “Please notice that the bonus check for Iris Jones was placed on top of Julia Jones’ performance report and vice versa” (Spanish: “Favor notar que el cheque para Iris Jones fue puesto sobre el reporte de desempeño de Julia Jones y vice-versa”). Thus, the script for the video is descriptive, and was translated word by word. We should also note that the manipulation was merely the language, not the syntax.
In order to ensure that the event was seen as unfair, participants were instructed that once the checks were filled out there was no way to change the amounts paid to the employees. Because the mistake is meant to be an accident, many of the participants might have missed the mistake if they were not looking for it closely. Thus, participants were also instructed to pay particular attention to information that represents our manipulation of an unfair mistake and were asked to watch the video twice. Participants were also asked to indicate the performance rating and the amount paid to Julia and Iris immediately after viewing the video the second time to ensure that they were able to identify the payment discrepancy. Participants were then asked to briefly describe, “what happened”/ “¿qué pasó?” in the video. Participants were then asked to respond to a brief questionnaire.
Measures
Participants were asked to imagine that they worked for the manager in the video when responding to all questions.
Agentive language
The first and second author coded all responses to the “what happened” question for the use of agentive versus non-agentive language. This construct was coded as 1 if the participant suggested that the manager was in some way responsible for the mistake (e.g., “The manager made a mistake”, “The manager switched the checks”) and as 0 if the participant used non-agentive language (e.g., “The checks were switched”, “a mistake was made”). Both coders reached adequate levels of agreement, ICC = .86; Krippendorff’s alpha = .86, and all discrepancies were discussed until agreement was reached. Footnote 3
Fairness perceptions
Fairness perceptions were measured using a three-item scale adapted from Ambrose and Schminke’s (Reference Ambrose and Schminke2009) overall justice scale (α = .93). An example item was “Overall, Iris Jones and Julia Jones were treated fairly in this situation” (1 = strongly disagree; 7 = strongly agree).
Behavioral intentions
Behavioral intentions to hold the responsible party accountable were measured using Williams and Anderson (Reference Williams and Anderson1991) OCBI scale to assess the intentions to engage in OCBs directed at the manager (α = .89). A sample item is “I would help the manager in the video if he had a heavy workload” (1 = strongly disagree; 7 = strongly agree). Lower intentions to engage in OCBs directed toward the manager was operationalized as indicating higher intentions to hold the responsible party accountable.
Results and Discussion
We first performed a MANOVA with agentive syntax, fairness perceptions, and behavioral intentions as the dependent variables and native language as the independent variable. The results revealed a significant main effect of language on agentive syntax F(1, 68) = 18.17, p < .01, η2 = .21, fairness F(1, 68) = 20.72, p < .001, η2 = .23, and behavioral intentions F(1, 68) = 4.26, p < .05, η2 = .06. Table 2 presents the means and correlations for each of the three dependent variables. Next, we directly tested the relationships hypothesized by Hypothesis 3. Specifically, we tested whether Language (i.e., Spanish versus English) led to the use of agentive versus non-agentive syntax, which was related to fairness perceptions and OCBI intentions. We used bootstrapping procedures, consistent with Preacher and Hayes (Reference Preacher and Hayes2008), to analyze the mediation hypotheses. The results revealed that the 95% confidence interval for the indirect effects of native language (Spanish versus English) on fairness perceptions via agentive versus non-agentive syntax, and subsequent behavioral intentions to hold the responsible party accountable by performing fewer OCBs did not include zero (Unstandardized Beta = –.26; S.E. = .18; p < .05 95% CI: –.55; –.12). Thus, Hypotheses 2 was supported Footnote 4 .
Table 2. Means and Correlations for Experiment 2
Note: *represents significance at p < .05; **represents significance at p < .01. For Language, 0 = Spanish; 1 = English. For Agentive Syntax, 0 = Non-Agentive; 1 = Agentive.
EXPERIMENT 3
Method
Participants
Experiment 3 tests Hypotheses 3, while controlling for cross-cultural values and native language. Specifically, we predicted that due to the non-agentive language used to encode unfair mistakes, Spanish speakers would be less likely to use entity fairness when reacting to an unfair mistake. A sample of 101 bilingual English-Spanish speakers (age = 29.55; % male = 66) participated in the experiment. Participants were bilinguals recruited via Amazon’s Mechanical Turk service (https://www.mturk.com). English and Spanish advertisements were set up to solicit bilingual speakers and the scope of the advertisement was limited to individuals whose IP addresses indicated that they were currently within in the United States.
Design and procedures
The experiment employed a 2 (Language: Spanish vs. English) x 2 (Entity Fairness: Fair vs. Unfair manager) between subjects design. Were randomly assigned to one of the four conditions. See Table 3 for descriptive for each condition. Participants All participants were instructed that they would watch a video about a hypothetical company. Participants read and listened to instructions in either English or Spanish. Consistent with the methods used in Experiment 2, instructions were developed by the first author, translated into Spanish by the second author who is a native Spanish speaker but is bilingual, and verified by an independent bilingual, native English speaker. The manipulation of the unfair mistake was identical to the manipulation used in Experiment 2. Specifically, participants were shown a scene in which a manager accidentally places the wrong check on each performance report resulting in an unfair distribution of bonuses.
Table 3. Means for Demographic variables in Experiment 3
Prior to viewing this event fairness scene, participants watched a short clip that included two scenes that manipulated entity fairness (see manipulations below). Participants were asked to pay close attention to the video and were instructed to watch the video twice before proceeding. After watching the video, participants were asked to briefly describe what happened in the video. Responses were content coded for the use of agentive vs. non-agentive language. Participants were then led to a brief questionnaire that asked participants “what happened, assessed fairness perceptions, intentions to engage in OCBs that would benefit the manager. Participants were also asked to complete measures of the cross-cultural values commonly studied.
Manipulation
Entity Fairness was manipulated in a manner consistent with the findings of Hollensbe, Khazanchi, and Masterson (Reference Hollensbe, Khazanchi and Masterson2008). These authors asked new entrants to describe how they assess the fairness of their supervisor. The content analysis revealed several supervisory characteristics that influence perceptions that a supervisor is fair (e.g., developmental behaviors. freedom in scheduling work, general positive and negative perceptions about the supervisor, etc.). Using these descriptions, we created two video scenarios suggesting the manager was generally fair or unfair. The actions in the video were silent, but a voice over accompanied the video to provide more contexts as well as to emphasize entity fairness. The same actors were used as the manager and employees in both scenarios. In order to ensure that participants only thought the entity was generally unfair and not biased toward one employee or the other, we instructed participants several times that the employees in the video were not the same employees receiving bonuses.
In order to further ensure that the manipulation of Entity Fairness was appropriate and that language (i.e., Spanish versus English) did not influence how fair the manager was perceived in each scenario, we conducted a pretest with 49 English and 37 Spanish speakers. All participants were solicited via mTurk. The only difference between the pretest and Experiment 3 is that the participants did not see the second half of the video in which the unfair mistake occurs, and participants only responded to a measure of entity fairness developed by Byrne and Miller (Reference Byrne and Miller2009; α = .99). The results of a one way ANOVA entering the measure of entity fairness perceptions as the dependent variable and the manipulation of entity fairness, native language, and the interaction term as the independent variables revealed a significant main effect of entity fairness F(1, 82) = 289.61, p < .01, no main effect of language F(1, 82) = .54, p = .47, and no interaction effect F(1, 82) = .63, p = .43. Thus, we were fairly confident that the entity fairness manipulation was appropriate and that language did not influence perceptions of entity fairness across conditions.
We should note that this manipulation of entity fairness is consistent with the organizational justice literature. In this literature, it is well-established that “fairness treatment” is part of the fairness construct. For example, in Colquitt (Reference Colquitt2001) it was conducted confirmatory factor analyses on organizational justice and found that 4 factors represented the best solution to fit their data. This implied distributive, procedural informational, and interpersonal justice dimensions. This last dimension is the degree to which managers are respectful and polite. For example, Skarlicki and Folger (Reference Skarlicki and Folger1997) used the term “interpersonally fair” to managers who are respectful and polite in their treatment to subordinates. As such, we do believe that our manipulation, using a manager that seems more or less polite, is affecting “entity fairness.”
Recall that we expected that the use of non-agentive language when encoding the fairness event would attenuate the use of entity fairness information. We did not expect that participants would differ in their perceptions of entity fairness, but instead that they would not use this information when the justice violation is encoded using non-agentive language because the responsible party is presumed to be less salient in this case.
Fair manager scenario
In scene 1 the fair manager crosses paths with an employee smiles, shakes her hand and has a brief conversation demonstrating that the manager is generally kind. In scene 2 the fair manager receives an award for his mentoring work, demonstrating that he is a supportive manager. The voiceover for this section was as follows: “Steve, a manager at Bramber, is described as accommodating and developmental towards employees. He allows his employees freedom in scheduling work arrangements and doing the job, and he has many positive characteristics. In fact, most employees describe Steve as a kind and fair manager. Steve was even recognized with an award for the mentoring work he has done with his employees over the years.” The English version of this video can be viewed at (http://lofvideos.participantpool.com/?ver=Study3EntityFairEnglish and the Spanish version can be viewed at http://lofvideos.participantpool.com/?ver=Study3EntityFairSpanish).
Unfair manager scenario
In scene 1, the unfair manager crosses paths with an employee and pushes past her despite the fact that she tentatively puts her hand out to shake his hand. In scene 2, the unfair manager shows favoritism toward one employee over the other. The voiceover for this section was as follows: “Steve, a manager at Bramber, is described as unwilling to accommodate or develop employees. He never allows his employees freedom in scheduling work arrangements or doing their job, and he has many negative characteristics. In fact, most employees describe Steve as a rude and unfair manager. Steve also frequently shows favoritism towards certain employees, while ignoring others.” The English version of this video can be viewed at (http://lofvideos.participantpool.com/?ver=Study3EntityUnfairEnglish and the Spanish version can be viewed at http://lofvideos.participantpool.com/?ver=Study3EntityUnfairSpanish).
Measures
Participants were asked to imagine that they worked for the manager in the video when responding to all questions. The measures of agentive language (ICC = .87; Krippendorff’s α =.77), and OCB intentions (α = .93) were identical to those used in Experiment 2.
Native language and language fluency
Native language was operationalized as a single item measures that asked participants to indicate whether English or Spanish was their primary language. Language fluency was measured using a single item measure of language fluency on the following scale, 1 = “I am really only fluent in Spanish”; 2 = “I am more fluent in Spanish, but I understand some English”; 3 = “I am more fluent in Spanish, but I understand most English”; 4 = “I am equally fluent in Spanish and English”; 5 = “I am more fluent in English, but I understand most Spanish”; 6 =“I am more fluent in English, but I understand some Spanish”; 7 =“I am really only fluent in English”. Native language and language fluency were used as controls in all analyses to ensure that the predicted effects were the result of the language used to encode the justice violation rather than the participant’s native tongue.
Trust
In order to be consistent with Choi (Reference Choi2008), we used the same trust measure derived from Roberts and O’Reilly (Reference Roberts and O’Reilly1974) trust in manager scale (α = .87). A sample item is, “If you were an employee at Bramber, how free would you feel to discuss the problems and difficulties in your job with the manager in the video without jeopardizing your position or having it held against you later?” The respondents assessed each item on a 7-point scale ranging from 1 (not at all) to 7 (very much).
Cross cultural values
Power distance, individualism, masculinity, uncertainty avoidance, and long-term orientation were measured using the procedures outlined by Hofstede, Hofstede, and Minkov (Reference Hofstede, Hofstede and Minkov2010). The measures of cultural values were used as a control in all analyses to ensure that cross-linguistic differences in the language used to encode the justice violation rather than other cultural norms were responsible for the hypothesized effects.
Results and Discussion
All analyses controlled for native language, and cross-cultural values. We first analyzed an omnibus test of our predictions using a two-way MANOVA in which OCB and agentive language were entered as the dependent variables and the manipulations of Entity Fairness, Language, and the interaction term were entered as the independent variables. Consistent with our predictions, the results revealed a significant main effect of native languages on the use of agentive language F(1, 90) = 10.44, p < .01, η2 = .10, and a significant interaction effect of entity fairness and native language on OCB Intentions F(1, 90) = 8.14, p < .01, η2 = .08. Participants in the Spanish condition used non-agentive language 49% of the time, whereas participants in the English condition only used non-agentive language 13% of the time when encoding the unfair event. Moreover, as shown in Figures 2 and 3, the results of the MANOVA support the predictions of Hypothesis 4, which suggested that the relationship between entity fairness and trust in the manager and OCB would be moderated by one’s native language. Specifically, the relationships were stronger for English speakers than Spanish speakers (See Table 4 and Figure 2). Furthermore, the form of the interaction is interesting. In Experiment 1, we found that Spanish speakers demonstrate less severe negative reactions to accidental justice violations. However, although not significant, compared to Spanish speakers, English speakers were more likely to indicate intentions to help the responsible social entity when the entity was presented as a fair manager. This pattern was reversed in the unfair entity condition. These results suggest that identifying the responsible party can have positive effects on reactions to justice violations encoded in English compared to Spanish, if the responsible party is generally fair. Although not explicitly hypothesized, we suggest that these effects occurred because encoding the accidental justice violations in Spanish attenuates the positive as well as the negative effects of entity fairness information.
Figure 2. Language by entity fairness on OCB intentions. Note. Error bars represent standard errors.
Figure 3. Agentive syntax by entity fairness on OCB Intentions. Note. Error bars represent standard errors.
Table 4. MANOVA Results for Experiment 3
Note: OCBI R 2 = .21; Agentive Syntax R 2 = .10.
Supplementary analysis
Although no specific predictions were made about the direct influence of agentive versus non-agentive syntax as a moderator of the relationship between entity fairness and OCB intentions, it was implied that the use of non-agentive language, rather than native language, is the proximal moderator of this relationship. If this perspective is correct, then the interaction effect of entity fairness and agentive versus non-agentive language on OCB intentions should be stronger than the interaction effect of entity fairness and native language. Unfortunately, no analytic methods have been developed to directly test the full model. Nonetheless, we preformed a supplemental analysis to test whether the full model is plausible. Specifically, we ran an additional MANOVA in which agentive versus non-agentive language and entity fairness were entered as the independent variables and OCB intentions were entered as the dependent variables. Consistent with our predictions, the results revealed a significant interaction effect of entity fairness and use of agentive versus non-agentive language on OCB intentions F(1, 90) = 3.91, p < .05, η2 = .04. (See Table 5 and Figure 3).
Table 5. ANOVA Results for Experiment 3 Supplementary Analysis
Note: OCBI R 2 = .25.
General Discussion
To our knowledge, this article represents the first research to explicitly address the influence of cross-linguistic differences on workplace criteria. Although we focus on how Spanish versus English linguistic norms influence reactions to unfair treatment, Evans and Levinson (Reference Evans and Levinson2009) suggest that there is a plethora of “phonetic, phonological, morphological, syntactic, and semantic” (p. 430) cross-linguistic differences that have been under-investigated by behavioral scientists. Thus, the research conducted herein not only adds to the literature on organizational justice, but also suggest that future research on how cross-linguistic differences influence perceptions and behaviors in the workplace might have important practical and theoretical implications for other areas of organizational behavior.
Beyond this general contribution, the results of each experiment extend the extant organizational justice research in several ways. First, the results of Experiment 1 extend research examining the positive effects of explanations on reactions to justice violations by introducing non-agentive syntax as a powerful rhetoric device for attenuating perceptions that the violation was unfair and subsequent behavioral intentions to hold the responsible party accountable. These results suggest that practitioners should consider their words carefully when describing justice violations. In addition, Experiment 1 provides preliminary evidence that cross-linguistic variation in the use of agentive versus non-agentive syntax when accidental justice violations are encoded using English versus Spanish represents the cognitive mechanism underlying the results of Experiments 2 and 3. Specifically, the use of non-agentive language when communicating incidents of unfairness assuaged negative reactions to justice violations even when the observer read the description in English – a language that typically encodes accidents using agentive syntax.
Second, the results of Experiment 2 extend cross-cultural organizational justice research by highlighting the influence of cross-linguistic differences in the use of agentive versus non-agentive syntax on fairness perceptions and subsequent intentions to hold the responsible party accountable for accidental justice violations. Specifically, cross-linguistic variation in the use of agentive versus non-agentive syntax when encoding the violation caused native Spanish speakers to perceive accidental justice violations as less unfair compared to English speakers. In turn, Spanish speakers were less likely to indicate that they would hold the responsible party accountable for the justice violations. These results suggest that practitioners should be cognizant that Spanish and English speakers who witness an accidental justice violations might react differently due to differences in cross-linguistic norms.
Third, the results of Experiment 3 extend research investigating the relationship between event and entity fairness perceptions on reactions to justice violations. Specifically, Experiment 3 found that when the justice violation was encoded using Spanish the influence of entity fairness information on OCB intentions was assuaged relative to when the violation was encoded in English. These results qualify previous findings by Choi (Reference Choi2008) by demonstrating that linguistic norms represent a boundary condition for the influence of entity fairness on reactions to accidental justice violations. In addition, the results of Experiment 3 were significant even when controlling for native language, language fluency, and cultural values. Therefore, these results provide further evidence that cross-linguistic differences in the norms used to encode accidental justice violations rather than native language or cross-cultural values are responsible for the findings in Experiment 3.
Fourth, the form of the interaction between entity fairness and language on subsequent reactions provide additional insights into the relationships between identifying a responsible party and reactions to justice violations. As shown in Figure 2, entity fairness information not only had a stronger influence on OCB intentions for bilingual participants who encoded the justice violation in English, but also caused these participants respond more negatively to an unfair manager and more positively to a fair manager than Spanish speakers. These results echo Choi’s (Reference Choi2008) suggestion that identification of a social entity can have a positive influence on reactions to unfair events, so long as the entity has a reputation as a generally fair manager. Interestingly, these findings contradict the general assertion by organizational justice theories that identifying the party responsible for justice violations exacerbate negative reactions (e.g., Folger & Cropanzano, Reference Folger, Cropanzano, Greenberg and Cropanzano2001). In addition, these results suggest that the use of non-agentive syntax as a rhetorical device should be considered with care. In particular, when the entity is generally fair agentive syntax might have a better chance of attenuating negative reactions to justice violations. However, these predictions require future research and scrutiny. Finally, the results of Experiment 3 suggest that managers would be wise to develop fair reputations, but they should not to rely on previously established reputations when mistakes lead to justice violations observed by Spanish speaking employees. Unlike English speaking employees that might give the fair, but bumbling manager the benefit of the doubt, Spanish-speaking employees might be less likely to let these justice violations slide.
Although this research provides several contributions, it is not without limitations. First, the generalizability of this research is somewhat limited. For one, the current article employed vignettes and asked participants to indicate behavioral intentions rather than examining actual behaviors. Although these methods seem appropriate for an initial investigation of the proposed phenomenon, future research should investigate these effects in field settings using actual behaviors. In addition, the participants in all studies were native English and/or Spanish speakers currently residing in the United States. However, similar effects should be demonstrated in any population whose language demonstrates similar linguistic norms when encoding accidental justice violations (i.e., the use of non-agentive syntax). The fact that (1) the results of Experiment 3 occurred even after controlling for cross-cultural values; (2) the results of Experiment 2 were mediated by the use of agentive vs. non-agentive language; and (3) English speakers were affected by the use of agentive vs. non-agentive syntax in Experiment 1 seems to lend some support to the potential generalizability of these effects to other languages and cultures.
Second, as a reviewer noted, our model is somewhat ambitious. It is difficult to establish the causality of fairness perceptions and OCBI intentions (reverse causality). However, we do believe that we have evidence for making a good claim on the causal of effects of language on fairness; and this is something we try to demonstrate in the paper. First, it is unlikely or impossible that fairness affects language in our studies, so we discard reverse causality. Second, we tried to discard another explanation, which would be the influence of a third variable. This is what we attempted to demonstrate in Experiment 3. In this experiment, we randomly assign participants to the language condition. Experiments and random assignment are the ideal method to assess causality (Angrist & Pischke, Reference Angrist and Pischke2008; Antonakis, Bendahan, Jacquart, & Lalive, Reference Antonakis, Bendahan, Jacquart and Lalive2010; Kausel, Reference Kausel2015; Shadish, Cook, & Campbell, Reference Shadish, Cook and Campbell2002). In addition, we control for the other explanation (the third variable) that could explain the effect s on fairness and OCBI.
Third, the research is limited to investigations of reactions to accidental justice violations. Nonetheless, several researchers have suggested that fairness is often ambiguous (e.g., Folger & Cropanzano, Reference Folger, Cropanzano, Greenberg and Cropanzano2001; Li et al., Reference Li, Evans, Christian, Gilliland, Kausel and Stein2011) and therefore justice violations might often seem accidental even when they occur through purposeful action. In addition, it is unlikely that managers are trying to act unjustly to their employees, thus fairness violations might frequently occur accidently. Thus, although the research is limited to accidental justice violations, we would expect that these situations are relatively common making the results applicable to a large portion of organizational justice events. However, more research in this area is necessary for further conclusions.
Fourth, Experiment 2 only used native Spanish and English speakers who interacted with the experiment in their native tongue. Thus, the effects of native language on fairness perceptions might have resulted from cross-cultural differences rather than cross-linguistic differences. However, Experiment 3 shows that even after controlling for cross-cultural values and native language, the language manipulation had an effect on fairness. Thus, it provides support to the idea that the results of Experiment 2 were caused by cross-linguistic, rather than cross-cultural differences. Moreover, the fact that agentive versus non-agentive syntax mediated the effects in Experiment 2 and were contingent on the language used to encode the justice violation rather than one’s native language in Experiment 3 provides more evidence in favor of the hypothesized model. Thus, the variation in agentive versus non-agentive syntax rather than cross-cultural differences seem to be responsible for the effects of Experiment 2.
This article explores the influence of language on reactions to unfair treatment. We find that non-agentive syntax represents a rhetoric device that can attenuate perceptions that a justice violation was unfair and subsequent intentions to hold the responsible party accountable. In addition, cross-linguistic differences between norms in Spanish and English that dictate the use of non-agentive versus agentive syntax when describing mistakes influences subsequent fairness perceptions, intentions to hold a responsible party accountable, and the use of entity fairness information in reactions to accidental justice violations. The studies reported herein suggest that the language used to encode unfair treatment can have a substantial influence on reactions to justice violations, and open up a number of promising directions for future research.
