Leader prohibitive voice shapes employee voice through psychological safety moderated by self-efficacy and generational differences

Abstract

This study examines how leader prohibitive voice behavior shapes employees’ own prohibitive voice under conditions of uncertainty. Drawing on Uncertainty Management Theory, we explore the mediating role of psychological safety and the moderating effects of self-efficacy and generational cohort. Using matched survey data from 302 leader–follower dyads in Chinese organizations, we tested a moderated mediation model. Results indicate that leader prohibitive voice promotes follower prohibitive voice both directly and indirectly via enhanced psychological safety. This indirect effect is stronger for employees with lower self-efficacy, who are more sensitive to contextual cues. Moreover, generational differences emerged: for Generation Z employees, leader modeling had a weaker effect, whereas psychological safety more strongly predicted their voice behavior compared to Millennials. These findings highlight that leaders reduce uncertainty not only through behavior modeling but also by fostering a psychologically safe climate, and that employee characteristics shape how such cues are interpreted. The study contributes to theory by integrating behavioral, cognitive, and generational perspectives on voice behavior and offers practical implications for leadership training, climate management, and generationally responsive organizational practices.

Introduction

In contemporary organizations marked by rapid technological change, intense competition, and rising uncertainty, proactively identifying and addressing potential risks has become increasingly critical¹. Prohibitive voice—employee communication aimed at highlighting harmful practices or problematic issues—is valued for enhancing organizational safety, adaptability, and innovation^2,3,4,5,6. However, prohibitive voice fundamentally differs from promotive voice not only in content but also in its psychological and relational consequences. Whereas promotive voice involves constructive suggestions for improvement and is generally perceived as forward-looking and opportunity-driven, prohibitive voice focuses on identifying existing problems or potential threats that could harm the organization^7,8. Because of its critical and risk-oriented nature, prohibitive voice is more likely to be interpreted as personal criticism or blame, often provoking defensiveness or discomfort in recipients—particularly when it challenges the status quo or implicates leaders’ decisions^9,10,11,12.

However, despite the rising relevance of prohibitive voice in uncertain organizational contexts, prior research has largely emphasized promotive voice—such as innovative suggestions or constructive ideas—while underexploring the unique antecedents and mechanisms that facilitate or hinder prohibitive voice^3,4,5,6,7,8. This limited attention is particularly problematic given that prohibitive voice entails greater psychological risk and ambiguity^5,7,13. Moreover, while previous studies have examined relational factors that promote general voice behavior^2,6,14 relatively few have investigated how employees interpret leaders’ signals when deciding whether to speak up about potential harm^10,12. As such, there is a need to identify when and how employees engage in prohibitive voice under uncertainty, and what leader behaviors reduce the perceived risks associated with it.

Given these dynamics, employees may hesitate to engage in prohibitive voice, not because it lacks value, but because it carries significant interpersonal and political risks. Despite its potential to enhance organizational safety and effectiveness, prohibitive voice can lead to negative consequences for the speaker, including managerial retaliation, strained workplace relationships, or exclusion from important decision-making processes^13,14,15,16. This creates a fundamental dilemma: employees must weigh the organizational benefits of voicing concerns against the personal costs of doing so, making the act of speaking up especially complex and uncertain.

Within organizations, leaders serve as key sources of these cues, shaping perceptions of which behaviors are acceptable and valued. While past studies have examined the role of leadership in shaping employee voice, they have predominantly focused on promotive voice behaviors and voice-supportive leadership styles—such as servant, transformational, and inclusive leadership—which foster an open climate for upward communication^17,18,19. These studies generally emphasize proactive expressions aimed at organizational improvement, assuming relatively low interpersonal risk.

Even in research that addresses prohibitive voice, leader influence has predominantly been conceptualized using generalized relational constructs, such as relational identification or leader–member exchange (LMX). For example, MacMillan et al.²⁰ examined how high-quality LMX relationships can foster follower prohibitive voice, suggesting that strong relational ties provide psychological security that encourages employees to take interpersonal risks. Similarly, Liang et al.²¹ emphasized the mediating role of relational identification in promoting voice under transformational leadership. However, these studies typically treat leader influence as an affective or cognitive relational state, rather than as a set of observable behaviors that explicitly model risk-bearing actions in uncertain contexts. This leaves a gap in understanding how specific leader behaviors—such as expressing one’s own prohibitive voice—may function as a normative signal that reduces ambiguity for followers deciding whether to speak up.

To extend this line of inquiry, it is necessary to move beyond motivational alignment and consider how leader behavior functions as a contextual signal under uncertainty. While social learning theory explains how followers imitate leaders they identify with, it does not fully address how employees assess whether speaking up is safe or acceptable, especially in ambiguous or high-risk environments. In such settings, employees often experience psychological uncertainty regarding the social and political consequences of voicing concerns. Uncertainty Management Theory (UMT)²² offers a valuable complementary perspective by proposing that individuals actively seek environmental cues to interpret ambiguous situations and guide appropriate responses. From this standpoint, leader prohibitive voice can serve not only as a behavior to emulate, but also as a normative signal that clarifies whether voicing concerns is permissible and supported. When leaders themselves engage in prohibitive voice, they convey that such behavior is legitimate and valued within the organization—thereby reducing ambiguity and enhancing employees’ sense of psychological safety, defined as a shared belief that the work environment is safe for interpersonal risk-taking without fear of negative repercussions²³. This shift from modeling to signaling highlights the role of leaders not only as interpersonal exemplars, but also as institutional interpreters of acceptable conduct in uncertain organizational contexts.

While leader prohibitive voice may reduce contextual uncertainty and enhance psychological safety, employees are not equally influenced by such signals. Individual differences—particularly self-efficacy, defined as the belief in one’s capacity to handle challenging situations²⁴—shape how individuals interpret and respond to environmental cues. Employees with high self-efficacy are more likely to tolerate interpersonal risk and engage in voice behavior even without strong contextual reassurance. In contrast, employees with low self-efficacy are more attuned to environmental signals and more dependent on psychologically safe conditions to speak up. Accordingly, psychological safety may be especially critical for encouraging follower prohibitive voice among low self-efficacy individuals. Building on this rationale, we propose that self-efficacy moderates the indirect effect of leader prohibitive voice on follower prohibitive voice via psychological safety. Specifically, when self-efficacy is low, employees are more likely to rely on contextual cues—such as leader signaling and safety climate—to determine whether it is safe to speak up, thereby strengthening the mediating role of psychological safety. In contrast, when self-efficacy is high, individuals may draw more on internal confidence and less on contextual reassurance, making the mediating role of psychological safety less pronounced. In sum, self-efficacy serves as a boundary condition that shapes how psychological safety translates leader prohibitive voice into follower prohibitive voice behavior.

In addition to self-efficacy, generational cohort may shape how employees respond to leader signals and psychological safety in uncertain contexts. Drawing on UMT, we suggest that individuals from different generational backgrounds interpret environmental cues and regulate behavior differently when facing ambiguity. Generation Z (born 1996–2010), who came of age amid economic volatility and technological disruption, tend to report higher anxiety and stress in ambiguous situations and rely more heavily on contextual signals of safety and fairness^25,26. As such, psychological safety may play a particularly critical role in enabling their engagement in risk-laden behaviors such as prohibitive voice. Moreover, their emphasis on autonomy and psychological empowerment suggests that self-efficacy may exert a stronger influence on behavior within this cohort. By contrast, Millennials (born 1980–1995) entered the workforce under more stable and hierarchical conditions and may be more accustomed to top-down leader cues. Accordingly, they may be less reliant on contextual reassurance and more likely to interpret leader prohibitive voice as a clear behavioral model. These differences raise the possibility that the effects of leader behavior, psychological safety, and self-efficacy may vary across cohorts. Therefore, these generational patterns suggest that the direct, indirect, and moderated pathways linking leader behavior to employee voice are not uniform across cohorts.

Unlike prior studies that have primarily drawn on social learning theory to explain leader–follower voice transmission, this study advances the conversation by incorporating UMT to address the psychological ambiguity surrounding prohibitive voice in high-risk environments. This dual-theoretical perspective enables a more comprehensive understanding of not only how employees learn from leader behaviors, but also when and why such modeling translates into actual voice behavior. By integrating UMT with leader behavior modeling, self-efficacy, and generational differences, the study reveals that the effectiveness of leader prohibitive voice is contingent upon both contextual signals and individual-level factors. It moves beyond generalized leadership styles to highlight the signaling role of specific leader actions, while also uncovering how psychological safety and self-efficacy interact differently across generational cohorts. These insights contribute to a more differentiated theory of employee voice that accounts for personal, situational, and generational contingencies. The findings offer important implications for designing psychologically safe work environments and leadership practices that are responsive to the diverse needs of today’s multigenerational workforce—particularly in contexts marked by uncertainty, interpersonal sensitivity, and organizational risk.

Literature review and research hypotheses

Voice behavior

Voice behavior is a discretionary and proactive communication process in which employees express ideas, suggestions, or concerns aimed at improving organizational effectiveness^27,28. It is broadly classified into two types: promotive voice and prohibitive voice⁷. While both contribute to organizational improvement, they differ in several aspects. Promotive voice is constructive and future-oriented, focusing on suggestions for enhancing organizational functions and performance, whereas prohibitive voice is preventive and risk-focused, aiming to identify and mitigate harmful organizational factors such as unethical practices, inefficiencies, or safety hazards. Promotive voice is forward-looking, emphasizing innovation and change for future improvement, while prohibitive voice is both retrospective and prospective, addressing past or potential future issues that may negatively impact the organization. Promotive voice generally carries lower interpersonal and professional risks, as it is perceived as contributing positively to organizational growth, whereas prohibitive voice involves higher risks, including potential relational damage, managerial resistance, or negative repercussions, as it often challenges the status quo^7,8,29,30.

Additionally, promotive voice is typically perceived favorably within organizations as it aligns with proactive problem-solving and business development, whereas prohibitive voice may be viewed more negatively, as it involves criticism of existing practices and can be perceived as disruptive. Despite its challenges, prohibitive voice is essential for risk mitigation, ethical governance, and long-term organizational sustainability^8,30. However, its inherent risks create dilemmas for employees, influencing whether and how they choose to engage in such behavior.

Uncertainty management theory and the prohibitive voice dilemma

Despite its organizational benefits, engaging in prohibitive voice is often fraught with interpersonal and reputational risks. Employees who express concerns about harmful practices may be perceived as disruptive, face resistance from colleagues and supervisors, or experience negative career consequences^13,15. This dilemma creates an internal conflict: while prohibitive voice behavior could prevent organizational failure and promote ethical behavior, it may also expose employees to backlash or social exclusion^31,32,33. Thus, employees must constantly assess whether voicing concerns is worth the potential risks.

UMT provides a valuable framework for understanding how employees navigate the risks associated with prohibitive voice²². Uncertainty is defined as a state in which individuals cannot accurately predict future outcomes or the consequences of their actions. In organizational contexts, uncertainty can arise when performance evaluation criteria are unclear, organizational policies are inconsistent, or employees are unsure how their actions will be perceived by colleagues and supervisors³⁴.

Prohibitive voice inherently aligns with UMT’s concept of uncertainty because employees who engage in it face ambiguity regarding potential outcomes. Raising concerns about organizational risks or unethical practices can lead to both positive consequences, such as organizational improvements and ethical corrections, and negative repercussions, including managerial retaliation or damaged workplace relationships. The uncertainty regarding whether their voice will be welcomed or punished discourages employees from prohibitive voice behavior. Given that employees often weigh the risks and benefits of engaging in prohibitive voice, they seek environmental cues to reduce uncertainty before deciding to express their concerns.

According to UMT, when individuals face uncertainty, they rely on social cues and organizational norms to guide their behavior³⁵. In the context of prohibitive voice, the clarity of leadership signals regarding whether such behaviors are acceptable plays a crucial role in employees’ decision-making. Without clear guidance, employees may hesitate, fearing unintended consequences. Thus, reducing uncertainty is essential for fostering an environment where employees feel psychologically safe to voice their concerns.

The role of leader prohibitive voice behavior in reducing uncertainty

Leaders play a critical role in shaping employees’ perceptions of uncertainty by signaling which behaviors are acceptable and encouraged^1,3,14,36. Numerous studies have highlighted that leadership significantly influences employee voice behaviors, as employees often base their decisions to speak up on their immediate supervisor’s behavior^{2,13,37,38,39}. Research suggests that because leaders occupy a hierarchical position with control over resources, rewards, and penalties, they directly shape employees’ perceptions of the risks associated with voicing concerns^36,37,40. Given that prohibitive voice involves questioning the status quo, employees tend to evaluate the potential consequences of their actions by “reading the wind”² to assess whether speaking up is safe.

Prior research has examined how different leadership styles, such as transformational, ethical, inclusive, and humble leadership, contribute to creating a supportive climate for voice behaviors^41,42,43. Transformational leaders encourage voice by fostering an environment of openness, intellectual stimulation, and individualized consideration^21,44,45,46. Ethical leaders, through their commitment to high moral standards, cultivate an environment where employees feel morally compelled and safe to express concerns^47,48. Humble leadership also plays a significant role in fostering prohibitive voice. Wang & Zhou⁴³ found that humble leaders—who acknowledge their limitations, appreciate employee contributions, and demonstrate openness to feedback—enhance psychological safety, thereby encouraging prohibitive voice behaviors. Humble leaders create an environment where employees feel secure in expressing concerns, reinforcing that their input is not only tolerated but essential for organizational improvement^49,50.

In contrast to these leadership styles, which primarily promote voice behaviors through relationship-building, direct leader behaviors—such as engaging in prohibitive voice—serve not only as social signals but also as behavioral models that provide employees with concrete guidance on how to navigate uncertainty. According to Social Learning Theory⁵¹ individuals learn behaviors by observing role models, especially those in positions of authority. When leaders visibly engage in prohibitive voice, employees not only infer that such behavior is safe (as per UMT), but also acquire a clear behavioral script for how and when to voice concerns. This dual function reinforces the acceptability of dissent and reduces the ambiguity surrounding voice behavior. In doing so, leaders establish strong normative expectations that speaking up is not only permissible but valued. This modeling process is particularly salient in high-uncertainty contexts, where observational learning substitutes for explicit instructions or policies.

Hypothesis 1

Leader prohibitive voice behavior is positively related to employee prohibitive voice behavior.

The role of psychological safety as a mediator

From a UMT perspective, psychological safety serves as a critical environmental mechanism that mitigates the uncertainty employees face when deciding whether to engage in prohibitive voice behavior. Psychological safety, defined as the shared belief that expressing one’s concerns will not result in negative interpersonal consequences²³ is essential for fostering open communication in organizations. In environments characterized by uncertainty, psychological safety provides employees with a sense of predictability and assurance that their concerns will not lead to punitive repercussions or damaged workplace relationships^49,52. This assurance reduces perceived ambiguity and fear associated with engaging in prohibitive voice behaviors.

When leaders model prohibitive voice, they reduce uncertainty not only at the individual level but also at the collective level by reinforcing an organizational norm that speaking up is both valued and safe. Leaders who explicitly engage in prohibitive voice demonstrate that raising concerns is a legitimate and constructive practice rather than a risky or unwelcome behavior. As a result, employees internalize these signals, leading to heightened psychological safety. Employees who perceive a psychologically safe environment experience lower perceived interpersonal risks, making them more likely to engage in prohibitive voice^53,54.

This approach shifts the focus beyond dyadic leader-follower relationships and emphasizes the organizational climate-building role of leadership in reducing uncertainty. Unlike prior studies that predominantly examined leader-member exchange (LMX)⁶ or leader identification⁵⁵ as mediating mechanisms, this study argues that psychological safety functions as a key mediator by transforming prohibitive voice from a personally risky action into a socially accepted and encouraged norm. Therefore, psychological safety serves as a key mechanism linking leader prohibitive voice to employee prohibitive voice by reducing uncertainty and fostering a climate where employees feel secure in expressing concerns.

Hypothesis 2

The positive relationship between leader prohibitive voice behavior and employee prohibitive voice behavior is mediated by psychological safety.

Self-efficacy as a key individual difference in uncertainty management

Building on the foundational logic of UMT, we further examine the conditions under which the psychological safety fostered by leader prohibitive voice is more or less likely to translate into employee voice behavior. According to UMT, individuals rely on contextual cues to assess the potential consequences of their actions under ambiguous conditions. Yet, the strength of these cues depends significantly on individual-level factors—most notably, self-efficacy.

UMT posits that when individuals face uncertain environments, they seek clarity through external signals to regulate their behavioral decisions. Leader prohibitive voice, as a behavioral cue, reduces uncertainty by establishing that speaking up about problems is acceptable and valued. This reduction in perceived interpersonal risk fosters psychological safety, which encourages employees to engage in prohibitive voice. However, this process is not uniform across all employees. Those with lower self-efficacy—defined as the belief in one’s capability to perform challenging tasks⁵⁶—are more likely to rely on such contextual signals to feel secure in voicing concerns. Without confidence in their own ability to navigate social or professional repercussions, these employees depend more on environmental assurances such as psychological safety to engage in voice behavior.

Conversely, individuals with high self-efficacy are more autonomous in their decision-making. They possess a robust sense of personal agency and are therefore less affected by environmental uncertainty. Even in the absence of strong psychological safety, they may still engage in prohibitive voice due to their confidence in managing interpersonal dynamics. This implies that psychological safety has a diminished role for high self-efficacy individuals, weakening the indirect effect of leader prohibitive voice.

Recent empirical findings lend support to this conceptualization. For example, Qian et al.⁵⁷ found that employees with low self-efficacy were more sensitive to contextual cues such as supervisor support and psychological safety, and this sensitivity significantly shaped their willingness to engage in voice behaviors. Similarly, Hao et al.⁵⁸ demonstrated that self-efficacy moderated the indirect effect of trust in coworkers on knowledge sharing behavior via psychological safety, with stronger effects for individuals lower in efficacy. These findings are consistent with Bandura’s social cognitive theory, which underscores that individuals with lower self-efficacy rely more on external cues to guide behavior in uncertain environments. In parallel, our study proposes that employees with lower self-efficacy are particularly responsive to the psychological climate fostered by leader voice and thus more likely to be influenced by it.

In our model, leader prohibitive voice functions as an initial behavioral signal that shapes the immediate psychological climate of the team. When leaders speak up about risks and problems, they reduce contextual ambiguity and demonstrate that such behaviors are not only tolerated but valued. This behavior directly fosters psychological safety by indicating that raising concerns will not result in punitive consequences. For employees with low self-efficacy—who are more susceptible to uncertainty and less confident in managing potential interpersonal risks^59,60,61—this increase in psychological safety acts as a crucial external cue. It alleviates internal hesitation and enables them to perceive speaking up as a viable and safe option. In other words, for low self-efficacy individuals, leader prohibitive voice initiates a chain reaction that creates a safer environment (via psychological safety), which they are more likely to utilize as a behavioral guide, thereby increasing their own prohibitive voice behavior.

In contrast, high self-efficacy employees are more internally assured and less dependent on environmental validation⁶². While they may still perceive increased psychological safety from leader behavior, it does not significantly change their likelihood of speaking up because their baseline confidence allows them to engage in voice even under less safe conditions. Thus, the mediating role of psychological safety is more impactful for those low in self-efficacy, reinforcing the strength of the indirect relationship. Accordingly, self-efficacy serves as a boundary condition in this indirect relationship: when self-efficacy is low, psychological safety becomes a decisive enabling factor that strengthens the indirect effect; when self-efficacy is high, the need for such environmental reinforcement diminishes, thereby weakening the mediated path.

Hypothesis 3

Self-efficacy moderates the indirect effect of leader prohibitive voice on employee prohibitive voice via psychological safety, such that the indirect effect is stronger when self-efficacy is low and less pronounced when self-efficacy is high.

Generational differences in responses to uncertainty and leader influence

UMT suggests that individuals respond to uncertainty differently depending on their personal characteristics and the cues they receive from the environment²². Generational cohort is one such characteristic that meaningfully shapes how individuals interpret uncertainty and social cues in organizational settings. Prior research has shown that generational cohorts differ in their workplace values, attitudes toward authority, and expectations for leadership^63,64 which may influence how they engage in upward voice behaviors in response to uncertainty.

Millennials and Generation Z differ notably in how they perceive and respond to uncertainty. Empirical studies suggest that Generation Z employees report higher levels of anxiety and stress when facing ambiguity, compared to Millennials, due to their exposure to economic volatility and rapid technological disruption during formative years^25,26. As a result, they tend to place greater emphasis on environmental signals of safety and fairness^65,66,67 which aligns with UMT’s assertion that individuals are more reliant on external cues when internal certainty is lacking. Psychological safety, in particular, becomes a critical resource for Z-generation employees to overcome uncertainty and engage in risk-laden behaviors such as prohibitive voice.

In contrast, Millennials, who entered the workforce under more stable conditions and with greater exposure to hierarchical organizational structures, may be less sensitive to uncertainty and more accustomed to speaking up even without strong contextual reassurance⁶⁸. Consequently, the effect of leader prohibitive voice behavior as a social cue may resonate more strongly with Millennials, who are more attuned to top-down modeling processes, whereas Generation Z may rely more on broader contextual or peer-based cues²⁵.

Furthermore, Generation Z employees tend to place greater value on autonomy, personal development, and psychological empowerment⁶⁹ suggesting that their self-efficacy may play a stronger role in shaping behavioral decisions under uncertainty. Those with higher self-efficacy may feel more capable of acting despite unclear consequences, while those with lower self-efficacy may depend heavily on psychological safety to feel secure in speaking up. Thus, not only do generational differences influence how leader cues and psychological safety are interpreted, but they also shape the extent to which individual resources like self-efficacy facilitate voice behavior. Based on these theoretical assumptions, we propose the following three sub-hypotheses regarding generational differences in the voice mechanism.

Hypothesis 4a

The effect of leader prohibitive voice behavior on employee prohibitive voice behavior (Hypothesis 1) will be weaker for Generation Z employees than for Millennial employees.

Hypothesis 4b

The mediating effect of psychological safety on the relationship between leader prohibitive voice and employee prohibitive voice (Hypothesis 2) will be stronger for Generation Z employees than for Millennial employees.

Hypothesis 4c

The moderating effect of self-efficacy on the indirect relationship between leader prohibitive voice and employee prohibitive voice via psychological safety (Hypothesis 3) will be stronger for Generation Z employees than for Millennial employees.

Methods

Date collection and sample

This study used survey data collected from matched leader–employee pairs in China. A total of 56 companies from diverse industries—including manufacturing, finance, IT, services, and government—participated to enhance the generalizability of the findings. To ensure behavioral observability, data were collected from departments with direct leaders in general management roles who could serve as visible role models. Participants came from various job functions such as administration, sales, and production. A one-to-one matching procedure was employed to minimize common method bias, with surveys distributed separately to leaders and followers. Leaders assessed follower prohibitive voice behavior (the outcome variable), while followers reported on leader prohibitive voice behavior (independent variable), psychological safety (mediator), and self-efficacy (moderator), along with demographic information.

Prior to data collection, all participants received email briefings outlining the study’s purpose, data use, storage procedures, and confidentiality protections. Informed consent was obtained from organizational managers, and all respondents were assured of anonymity and the voluntary nature of their participation. Data were securely stored, and no personally identifiable information was collected. Only the designated researcher had access to the completed surveys. All research procedures were approved by the Ethics Committee of Suzhou City University (Approval code: SZCU12020620241201), and conducted in accordance with relevant guidelines and regulations. Informed consent was obtained from all participants, and research involving human participants was performed in accordance with the Declaration of Helsinki.

Of the 350 leader–follower dyads invited, 320 responses were received (91.4% response rate), with 302 valid dyads retained for final analysis after excluding mismatches and incomplete responses. Among followers, 47.0% were male and 53.0% were female. Age distribution was: 41.1% in their 20s, 49.0% in their 30s, and 9.9% in their 40s (SD = 6.39). In terms of education, 16.2% had a high school diploma, 22.8% attended two- or three-year colleges, 46.4% held four-year university degrees, and 14.6% had graduate-level education. Regarding tenure with their leader, 16.8% had less than one year of experience, 56.0% had one to five years, 18.9% had five to ten years, and 8.3% had over ten years. Generational classification identified 124 participants (41.1%) as Generation Z (born 1996–2010) and 178 (58.9%) as Millennials (born 1980–1995).

Measures

To test the hypotheses, a questionnaire was administered using a 5-point Likert scale, ranging from 1 = strongly disagree to 5 = strongly agree. As all scale items were originally developed in English, a standard back-translation procedure was employed to translate them into Chinese⁷⁰. The scales used in the questionnaire have demonstrated reliability and validity, as supported by previous studies.

Prohibitive voice behavior. Liang et al.’s⁷ five-item scale was used to measure both leader and follower prohibitive voice behavior. Sample items for follower prohibitive voice behavior are, “This follower dares to voice opinions on issues that might affect work unit efficiency, even if doing so would embarrass others.” To assess leader prohibitive voice behavior, the phrase “This follower” was replaced with “My leader” in all items. The Cronbach’s alpha for follower prohibitive voice behavior was 0.88, and for leader prohibitive voice behavior, it was 0.90. In line with conventional research standards, a Cronbach’s alpha of 0.70 or above is considered indicative of acceptable internal consistency.

Follower psychological safety. Edmondson’s²³ seven-item scale was used to assess followers’ psychological safety. Sample items are, “No one on this team would deliberately act in a way that undermines my efforts.” The scale demonstrated acceptable internal consistency, with a Cronbach’s alpha of 0.80.

Self-efficacy. We measured focal employees’ general self-efficacy with the eight item New General Self-Efficacy Scale⁷¹. Sample items are ,“When facing difficult tasks, I am certain that I will accomplish them” and “In general, I think that I can obtain outcomes that are important to me.” The Cronbach’s alpha for this scale was 0.89, indicating high internal consistency.

Generation. Generation was coded as 0 = Generation Z, and 1 = Millennials. This classification was based on widely accepted standards in Chinese demographic and academic research, which define Millennials as individuals born between 1980 and 1995, and Generation Z as those born between 1996 and 2010⁷².

Control variables. Four participant characteristics—gender, age, education, and tenure with the leader—were statistically controlled based on prior studies linking them to voice behavior^2,7,28,73. Gender was coded as 0 (male) and 1 (female); education was coded as 1 = high school, 2 = 2–3 year college, 3 = 4-year university, and 4 = graduate school. Age and tenure with the leader were reported in years. To isolate the effect of leader prohibitive voice, leader promotive voice behavior was controlled, as it may shape employees’ general perceptions of a voice-supportive climate. A five-item scale developed by Liang et al.⁷ was used for this variable. A sample item is: “My leader proactively voices constructive suggestions that help the unit reach its goals.” The Cronbach’s alpha for this scale was 0.89, indicating acceptable internal consistency.

Results

Confirmatory factory analysis

To assess the construct validity of the key variables used in this study, a confirmatory factor analysis (CFA) was conducted. As shown in Table 1, the hypothesized four-factor model—comprising leader prohibitive voice, psychological safety, self-efficacy, and follower prohibitive voice—demonstrated the best overall model fit compared to the alternative models (χ²[269] = 487.13, CFI = 0.94, IFI = 0.94, TLI = 0.94, RMSEA = 0.05).

Table 1 Comparison of measurement models.

In contrast, the three-factor models, which combined leader prohibitive voice with psychological safety (Model 1) or leader prohibitive voice with follower prohibitive voice (Model 2), exhibited significantly worse model fit (e.g., Model 2: Δχ² = 1741.82, p < 0.001, RMSEA = 0.09, CFI = 0.84). The two-factor model, which combined leader prohibitive voice, psychological safety, and self-efficacy into a single factor (Model 3), also showed inferior fit indices (Δχ² = 1986.13, p < 0.001, RMSEA = 0.10, CFI = 0.79). These results confirm that leader prohibitive voice, psychological safety, self-efficacy, and follower prohibitive voice are empirically distinct constructs, thereby supporting the discriminant validity of the measurement model and justifying the use of the four-factor structure in subsequent hypothesis testing.

In addition, to evaluate the potential threat of common method bias, Harman’s single-factor test was conducted using exploratory factor analysis (principal component analysis without rotation). The results revealed that the first unrotated factor accounted for 40.099% of the total variance, which is below the commonly recommended threshold of 50%⁷⁴. This suggests that common method bias is unlikely to significantly distort the results of this study. Therefore, the observed relationships among variables can be interpreted with greater confidence regarding their validity. The detailed variance explained by each factor is visualized in Fig. 1.

Fig. 1

Variance explained by each component in Harman’s single-factor test.

Descriptive statistics and correlation analysis

Table 2 presents the descriptive statistics and intercorrelations among the key variables used in this study.

Table 2 Comparison of measurement models.

Leader prohibitive voice, the independent variable, was positively and significantly correlated with both psychological safety (r = 0.46, p < 0.001), the mediating variable, and follower prohibitive voice (r = 0.54, p < 0.001), the dependent variable. Psychological safety also exhibited a significant positive correlation with follower prohibitive voice (r = 0.53, p < 0.001) and with self-efficacy (r = 0.54, p < 0.001), which served as the moderating variable. In contrast, generation showed no statistically significant correlations with any of the focal variables, including leader prohibitive voice (r = 0.06, ns.), psychological safety (r = − 0.02, ns.), self-efficacy (r = − 0.04, ns.), and follower prohibitive voice (r = 0.05, ns.). These correlations provide preliminary insights into the relationships among the study variables. The direction and strength of these associations were further examined through hierarchical regression and moderated mediation analyses to test the hypothesized model more rigorously.

Hypothesis testing

Prior to conducting hierarchical regression analyses for hypothesis testing, all continuous variables used in the regression models (excluding demographic controls) were mean-centered to minimize multicollinearity⁷⁵. The variance inflation factor (VIF) values were all below the conventional threshold, with the highest VIF recorded at 2.04, indicating that multicollinearity was not a concern (see VIF values in Table 3).

Table 3 Hierarchical regression results on follower prohibitive voice behavior.

Main effect. Hypothesis 1 predicted a positive relationship between leader prohibitive voice behavior and employee prohibitive voice behavior. As shown in Model 2 of Table 3, after controlling for demographic variables, leader prohibitive voice behavior was positively and significantly related to employee prohibitive voice behavior (β = 0.29, p < 0.001). Therefore, Hypothesis 1 was supported.

Mediating effect. Hypothesis 2 proposed that psychological safety mediates the relationship between leader prohibitive voice behavior and employee prohibitive voice behavior. To test this mediation effect, a bootstrapping procedure was conducted based on Preacher et al. (2007)⁷⁶. The bootstrap sample was set to 5000 iterations, and the indirect effect was estimated using bias-corrected 95% confidence intervals along with unstandardized coefficients and standard errors. As shown in Table 4, the indirect effect of psychological safety was 0.130, with a 95% confidence interval ranging from 0.070 to 0.212. Since the confidence interval did not include zero, Hypothesis 2 was supported.

Table 4 Direct and mediating path results.

Furthermore, the direct effect of leader prohibitive voice behavior on follower prohibitive voice behavior remained statistically significant (B = 0.237, 95% CI [0.091, 0.384]) even after including psychological safety in the model. This indicates that psychological safety serves as a partial mediator rather than a full mediator. Hence, while psychological safety helps explain how leader prohibitive voice facilitates follower prohibitive voice, leaders also influence voice behavior through other mechanisms beyond psychological safety.

Moderated mediation effect of self-efficacy. Hypothesis 3 proposed that the indirect effect of leader prohibitive voice on employee prohibitive voice through psychological safety would be moderated by self-efficacy, such that the indirect effect is stronger when self-efficacy is low and less pronounced when self-efficacy is high. To test this moderated mediation effect, a bootstrapping procedure was conducted following the approach of Preacher et al.⁷⁶ estimating conditional indirect effects at high (+ 1 SD) and low (–1 SD) levels of self-efficacy. As shown in Table 5, the indirect effect of leader prohibitive voice through psychological safety was significant when self-efficacy was low (point estimate = 0.14, 95% CI [0.07, 0.21]), but nonsignificant when self-efficacy was high (point estimate = 0.03, 95% CI [–0.03, 0.09]). Because the confidence interval for low self-efficacy excluded zero, whereas the interval for high self-efficacy included it, Hypothesis 3 was supported. The interaction pattern for the statistically significant moderating effect of self-efficacy was visualized in Fig. 2, following the procedure recommended by Aiken and West⁷⁵.

Table 5 Results of bootstrapped conditional indirect effects by self-efficacy level.

Fig. 2

Moderating effect of self-efficacy.

To further validate this interaction and address the limitation of using only arbitrary points in the simple slopes analysis—such as fixed values at ± 1 SD commonly used in the Aiken and West⁷⁵ approach—we conducted a Johnson–Neyman (J-N) analysis, which identifies the exact range of the moderator within which the focal effect is statistically significant. This method allows for a more nuanced and precise understanding of when the moderated mediation occurs by examining the full continuum of the moderator. As shown in Fig. 3, the Johnson–Neyman results revealed that psychological safety had a significant positive effect on employee prohibitive voice only when self-efficacy was below approximately 0.386 (centered). Above this threshold, the effect became statistically nonsignificant, suggesting that the indirect effect of leader prohibitive voice via psychological safety is particularly pronounced when employees possess lower self-efficacy. This corresponds to a raw self-efficacy score of approximately 2.82, indicating that the moderated mediation effect becomes statistically meaningful only when employees’ self-efficacy falls below this level.

Fig. 3

Johnson–Neyman plot.

Boundary conditions of generation. Hypothesis 4a predicted that the direct effect of leader prohibitive voice behavior on employee prohibitive voice behavior would be weaker for Generation Z employees than for Millennials. As shown in the results of A: Direct path (Psychological safety not included) in Table 4, the direct effect was statistically significant and positive for Millennials (effect = 0.299, 95% CI [0.138, 0.461]), but nonsignificant for Generation Z (effect = − 0.004, 95% CI [–0.383, 0.375]). This suggests that Millennials are more likely to align their own voice behavior with leader cues, whereas Generation Z employees may be less influenced by direct leader modeling. These results support Hypothesis 4a.

Hypothesis 4b predicted that the mediating effect of psychological safety (as proposed in Hypothesis 2) would be stronger among Generation Z employees compared to Millennials. The results support this prediction. As shown in the mediating path of Table 4, the indirect effect of leader prohibitive voice on employee prohibitive voice via psychological safety was significant for both groups, but notably stronger among Generation Z (effect = 0.228, 95% CI [0.056, 0.494]) than Millennials (effect = 0.100, 95% CI [0.043, 0.184]). Therefore, Hypothesis 4b was supported.

Hypothesis 4c posited that the moderating effect of self-efficacy on the indirect relationship between leader prohibitive voice behavior and employee prohibitive voice behavior via psychological safety (as posited in Hypothesis 3) would be stronger for Generation Z employees than for Millennials. The conditional indirect effects reported in Table 5 revealed that for Generation Z, the mediating effect was significant at the low level of self-efficacy (effect = 0.286, 95% CI [0.006, 0.634]), but not at the mean or high levels, suggesting that leader prohibitive voice particularly promotes voice behavior via psychological safety among low self-efficacy Generation Z employees. For Millennials, the conditional indirect effects were significant at both low (effect = 0.108, 95% CI [0.048, 0.203]) and mean levels (effect = 0.069, 95% CI [0.029, 0.140]), but not at the high level of self-efficacy. These findings indicate a more pronounced moderated mediation effect within the Generation Z subgroup under low self-efficacy conditions. Therefore, Hypothesis 4c was partially supported.

Post-hoc analysis. To further assess the generalizability of our hypothesized model across different forms of voice behavior, we conducted a post hoc analysis using follower promotive voice as the dependent variable. While the primary focus of this study was on follower prohibitive voice due to its stronger association with uncertainty and interpersonal risk, prior research suggests that promotive and prohibitive voice often co-occur and may share similar antecedents^7,8.

We replicated the hierarchical regression procedure presented in the main analysis using follower promotive voice behavior as the outcome (see Table 6). Results showed that leader promotive voice behavior was a strong and consistent predictor of follower promotive voice across all models, with standardized coefficients ranging from β = 0.27 to β = 0.42 (p < 0.001 in all models). In addition, leader prohibitive voice behavior significantly predicted follower promotive voice in the earlier models (β = 0.027, p < 0.001 in Model 2; β = 0.17, p < 0.01 in Model 3), but its effect became non-significant after psychological safety was included, suggesting a potential indirect influence through psychological mechanisms such as safety perception.

Table 6 Hierarchical regression results on follower promotive voice behavior.

Psychological safety significantly predicted follower promotive voice in Model 3 (β = 0.28, p < 0.001) and remained significant in Model 5 (β = 0.17, p < 0.001). Furthermore, the interaction term between psychological safety and self-efficacy (PS × SE) was also significant in Model 5 (β = − 0.16, p < 0.01), indicating that the effect of psychological safety on follower promotive voice was stronger for employees with lower self-efficacy.

Compared to follower promotive voice, the effects observed in the prohibitive voice model were generally stronger. In the prohibitive voice model (Table 3), the effect of psychological safety remained relatively robust even after controlling for self-efficacy (β = 0.19, p < 0.001 in Model 5), while the same path was slightly weaker in the promotive voice model (β = 0.17, p < 0.001 in Model 5, Table 6). Furthermore, although the PS × SE interaction coefficient was numerically larger in the promotive model (β = − 0.16 vs. − 0.09), it accounted for less additional explained variance in the outcome. Specifically, the R² change from Model 4 to Model 5 was 0.01 in the prohibitive voice model but 0.02 in the promotive model, which may initially appear contradictory. Yet, this highlights that in the prohibitive voice model—where uncertainty is higher—even a small R² increase reflects a meaningful contribution, as the model is more closely aligned with theoretical expectations regarding contextual cue salience.

Overall, while the proposed mechanism applies to both types of follower voice, the moderated mediation dynamics are more salient in follower prohibitive voice, where contextual ambiguity and interpersonal risk are higher. This is consistent with UMT, which emphasizes that contextual signals like leader behavior exert stronger influence in high-risk situations. Accordingly, our decision to focus the main theoretical model on follower prohibitive voice is both conceptually justified and empirically supported.

Discussion

Overall findings

To address the dilemma of prohibitive voice—balancing organizational benefit with interpersonal risk—this study adopted UMT²² as its framework. Using data from 302 matched leader–follower dyads in China, we found that leader prohibitive voice behavior encouraged follower voice both directly and indirectly via enhanced psychological safety. Self-efficacy moderated the indirect effect of leader prohibitive voice on follower prohibitive voice via psychological safety, such that the indirect effect was stronger among individuals with lower self-efficacy. Generational differences also emerged: among Generation Z employees, the direct influence of leader behavior was weaker, while the mediating role of psychological safety and the conditional effect of self-efficacy were more pronounced compared to Millennials. In a post-hoc analysis, we found that these psychological mechanisms were more salient for prohibitive voice than for promotive voice, reinforcing the theoretical focus on voice under uncertainty. These findings highlight complex mechanisms through which leader behavior, individual traits, and generational context jointly shape prohibitive voice. The following section elaborates on the theoretical and practical implications, limitations, and future research directions.

Theoretical implications

This study deepens the theoretical understanding of prohibitive voice by proposing and validating a multi-dimensional framework grounded in UMT²². By incorporating leader behavior, psychological mechanisms, individual traits, and generational context, it extends UMT across four key dimensions: behavioral cues, psychological processes, boundary conditions, and cohort sensitivity. Each hypothesis aligns with one of these dimensions, collectively offering a comprehensive view of how employees manage uncertainty in the context of prohibitive voice.

Leader prohibitive voice as a behavioral cue (Hypothesis 1). This study extends UMT²² to the domain of leader-initiated prohibitive voice behavior, offering a key theoretical contribution. UMT posits that in ambiguous or threatening contexts, individuals rely on salient cues to determine what is appropriate or safe. While past UMT research has focused on cues related to fairness, justice, or trust^77,78,79 little attention has been paid to leader communication, especially behaviors that involve expressing concerns or identifying risks.

By showing that leader prohibitive voice positively predicts follower prohibitive voice, this study identifies such behavior as a normative signal that legitimizes dissent. Despite its critical tone, leader prohibitive voice reduces ambiguity and clarifies expectations, encouraging employees to speak up. This finding complements prior work emphasizing promotive voice and supportive leadership^39,42 highlighting that even cautionary expressions have modeling value under uncertainty. In addition to serving as an uncertainty-reducing cue (UMT), leader prohibitive voice also functions as a behavioral model from the perspective of Social Learning Theory²⁴. Employees learn not only through formal instruction but also by observing others’ behavior and its consequences. When leaders openly express concerns about harmful practices, followers may interpret this as both a signal of acceptability (UMT) and an actionable example to emulate (SLT), reinforcing the legitimacy and enactment of prohibitive voice.

More broadly, the study positions leaders not just as procedural authorities but as behavioral exemplars. In situations where employees are unsure how their concerns will be perceived, leader prohibitive voice signals a psychologically safe climate for voicing potential risks. This advances UMT by emphasizing the behavioral—not only perceptual—nature of uncertainty cues, and illustrates how leader communication shapes implicit team norms around voice behavior.

Psychological mechanism of uncertainty reduction via psychological safety (Hypothesis 2). This study makes a second theoretical contribution by identifying psychological safety as the mechanism through which leader prohibitive voice reduces uncertainty. While previous studies have established psychological safety as a key antecedent of voice^23,80 this research positions it as a core UMT mechanism that translates external behavioral cues into follower decision-making.

Empirical results for Hypothesis 2 show that leader prohibitive voice enhances psychological safety, which in turn promotes follower prohibitive voice. This indicates that leader influence operates not only through visible modeling but also by shaping followers’ internal assessments of interpersonal risk. Psychological safety thus reflects an internalized sense of predictability, consistent with UMT’s premise that individuals seek cues to reduce ambiguity before acting⁸¹. This finding advances UMT by moving beyond a simple cue–response model, highlighting that uncertainty management involves cognitive and emotional appraisal. Rather than merely reacting to leader behavior, employees interpret its interpersonal implications—such as potential support or punishment.

By introducing psychological safety as a mediator, this study adds psychological depth to UMT and responds to calls for more cognitively enriched perspectives⁸². It also distinguishes this work from Social Cognitive Theory by emphasizing perceived safety—not just efficacy or modeling—as a prerequisite for voice. Ultimately, our results suggest that leader cues alone are insufficient; employees must also perceive the environment as psychologically safe. In this sense, psychological safety is not a byproduct of leadership, but a pivotal bridge connecting leader intentions to follower behavior, thereby integrating external and internal elements of uncertainty reduction.

Self-efficacy as an individual-level boundary condition (Hypothesis 3). This study’s third major contribution lies in identifying self-efficacy as an individual-level boundary condition that shapes how employees respond to uncertainty-reducing cues. Building on the behavioral (leader cues) and cognitive (psychological safety) layers of UMT, we introduce self-efficacy as a key moderating resource in the uncertainty reduction process. Our moderated mediation analysis shows that the indirect effect of leader prohibitive voice on follower voice—via psychological safety—is significantly stronger for employees with low self-efficacy. This suggests that individuals with lower confidence rely more on external signals to guide their behavior, while those with high self-efficacy may speak up regardless of leader behavior or perceived safety.

These findings extend UMT by showing that the effectiveness of uncertainty-reducing cues depends on the perceiver’s internal psychological state. Rather than assuming uniform cue responsiveness, we reveal that individuals’ reliance on cues varies with their coping capacity. Specifically, low self-efficacy increases sensitivity to leader signals and psychological safety, making these cues more impactful. This aligns with prior research⁸³ indicating that low self-efficacy employees are more attuned to external guidance under uncertainty. By foregrounding self-efficacy, our study enriches UMT by integrating person–context interactions and showing that cue utilization is filtered through self-perceived competence. This insight offers practical implications for tailoring leader communication to employees’ psychological needs, especially for those vulnerable to uncertainty.

Generational differences in uncertainty navigation (Hypothesis 4). This study’s fourth contribution extends UMT by incorporating generational differences into how employees navigate uncertainty. While UMT assumes that individuals rely on social cues—like leader behavior—to reduce uncertainty^14,84 our findings challenge this by showing that Generation Z and Millennials respond differently to such cues. Specifically, Hypothesis 4a revealed no significant direct relationship between leader prohibitive voice and follower voice among Gen Z employees, whereas Millennials showed a clear modeling effect. This suggests that Gen Z employees are less likely to view leader behavior as a normative guide for their own actions, diverging from core UMT assumptions. These results align with recent research indicating that Gen Z tends to be more autonomous, digitally oriented, and less influenced by hierarchical leadership cues^66,67,68. For instance, Wang et al. (2022)²⁵ found that Gen Z relies more on peer input or digital systems than traditional managerial modeling. Thus, conventional role-modeling may be less effective for this cohort.

However, Hypothesis 4b adds nuance: although direct modeling was absent, the indirect effect of leader prohibitive voice via psychological safety was stronger among Gen Z than Millennials. This implies that Gen Z interprets uncertainty primarily through perceived psychological climate rather than observable leader behavior. In other words, for Gen Z, psychological safety—not leader actions—serves as the key lens for interpreting uncertainty^23,80. When they feel safe, Gen Z employees are more likely to engage in prohibitive voice, even without strong behavioral cues. Hypothesis 4c further revealed that this indirect effect was strongest for Gen Z employees with low self-efficacy, whereas for Millennials, the effect was more stable across self-efficacy levels. This suggests that Gen Z employees with fewer internal resources are especially dependent on contextual support to overcome voice barriers, consistent with prior research⁸⁴.

Comparing follower promotive and prohibitive voice mechanisms (Post-hoc Analysis): To enhance the robustness and generalizability of our proposed model, we conducted a post-hoc analysis by substituting the outcome variable with follower promotive voice behavior. While our theoretical model primarily focused on follower prohibitive voice—given its higher levels of interpersonal risk and uncertainty—this additional analysis aimed to determine whether the hypothesized mechanisms would similarly apply to promotive voice behavior. As shown in Table 6, the results partially mirrored the findings from the prohibitive voice model. The main effect of psychological safety remained significant even after accounting for self-efficacy (β = 0.17, p < 0.001 in Model 5), although it was slightly weaker than in the prohibitive voice model (β = 0.19, p < 0.001). Interestingly, the interaction effect between psychological safety and self-efficacy (PS × SE) was also significant (β = − 0.16, p < 0.01), indicating that the proposed moderated mediation mechanism may extend to promotive voice as well. However, the R² change associated with this interaction (ΔR² = 0.02) was slightly greater than that of the prohibitive model (ΔR² = 0.01), suggesting nuanced differences in the salience of contextual cues.

These findings suggest that while our dual-process framework is applicable across both types of employee voice, the influence of contextual uncertainty and self-efficacy is more pronounced in follower prohibitive voice. This aligns with the theoretical reasoning that prohibitive voice, which involves pointing out risks or problems, is inherently more uncertain and interpersonally risky, thereby making employees more sensitive to contextual cues such as leader behavior and psychological safety. In contrast, promotive voice may require less uncertainty navigation, resulting in a comparatively attenuated role of self-efficacy and psychological safety interaction.

Collectively, these findings bridge Social Learning Theory and Uncertainty Management Theory to provide a dual-process explanation of how prohibitive voice behavior unfolds in organizations. While social learning captures how leader behaviors are modeled, UMT accounts for how employees interpret these behaviors under varying degrees of uncertainty. This integration offers a more comprehensive and psychologically grounded perspective on the dynamics of voice behavior—especially under risk—and extends prior theorizing by showing that imitation and uncertainty reduction jointly shape whether and how employees speak up.

Practical implications

This study offers several important practical implications for promoting prohibitive voice in organizations. First, the findings emphasize the critical role of leaders in resolving the psychological dilemma employees often face when considering whether to speak up about risks or concerns. When leaders themselves engage in prohibitive voice behavior, they send strong normative cues that such expressions are not only acceptable but expected. This modeling function is especially important under conditions of uncertainty, where employees seek guidance on what is appropriate behavior. Therefore, organizations should develop leadership training programs that go beyond abstract competencies and focus specifically on encouraging leaders to model voice behavior constructively, respond supportively to employee input, and create spaces where raising concerns is normalized.

Second, the study underscores the importance of cultivating a psychologically safe organizational climate. Psychological safety emerged as a key mechanism through which leader behavior influences employee voice, particularly for those with lower self-efficacy or those belonging to younger generations. This implies that organizations cannot rely solely on individual leaders to foster voice behavior; rather, they must make structural and cultural investments to institutionalize psychological safety. This includes designing feedback systems that reduce fear of punishment, encouraging inclusive dialogue across levels, and ensuring that voice behavior is treated as a valued form of contribution rather than defiance or disloyalty.

Third, from an individual perspective, the findings highlight that employees’ responses to uncertainty are not solely determined by their internal confidence but are also shaped by the environment in which they operate. Employees with lower self-efficacy are particularly sensitive to external cues, such as the presence of supportive leaders or inclusive team climates. As such, employees should be encouraged to build awareness of how they interpret their environment and to actively seek out or shape contexts that support safe communication. In addition, organizations can provide targeted support, such as mentoring or voice-skills development sessions, to help employees gain confidence in speaking up gradually and strategically.

Fourth, the study reveals important generational differences in how employees interpret leader behavior and environmental cues. Specifically, while Millennials tend to respond directly to leader modeling, Generation Z employees are more influenced by the overall psychological climate than by hierarchical authority. This suggests that traditional, top-down leadership approaches may be less effective for Gen Z, and that trust, autonomy, and peer-based feedback may be more critical for encouraging voice in this cohort. Leaders and organizations should therefore adapt their communication and influence strategies to reflect these generational preferences. At the same time, creating awareness among employees of different generations about these differences can help prevent misinterpretation and foster a more inclusive and empathetic workplace culture.

Lastly, our post-hoc analysis provided meaningful insight into how the same psychological mechanism—comprising psychological safety and self-efficacy—functions differently for promotive versus prohibitive voice. Given that prohibitive voice inherently involves greater interpersonal risk and ambiguity, the effects of contextual signals like leader support and psychological safety are more pronounced. In contrast, promotive voice is more aligned with employees’ personal motivations and growth orientation. These findings imply that a one-size-fits-all approach to voice promotion may be ineffective. Instead, organizations should tailor their leadership development, feedback systems, and communication strategies based on the specific type of voice they aim to foster.

Taken together, these practical implications suggest that effective promotion of prohibitive voice in organizations requires a multi-dimensional approach that aligns leadership behavior, organizational climate, individual support, and generational sensitivity. By integrating these elements, organizations can build a more responsive and psychologically secure environment where employees across diverse backgrounds and confidence levels feel empowered to speak up.

Limitations and future research

Despite its contributions, this study is not without limitations, which provide meaningful directions for future research. First, the cross-sectional nature of the data limits the ability to draw causal conclusions, particularly regarding the mediating role of psychological safety and the conditional indirect effect of self-efficacy. Although this study employed Harman’s single-factor test and confirmatory factor analysis to evaluate potential common method bias, we acknowledge that these statistical tests have limitations in fully capturing all sources of method variance. Notably, our data collection design involved matched leader–employee dyads, where leaders rated follower prohibitive voice behavior (the outcome variable), while employees reported on leader prohibitive voice behavior, psychological safety, and self-efficacy. This multi-source data collection approach reduces the likelihood of common method bias inflating the observed relationships. Nevertheless, we recognize that some residual bias may remain. Therefore, future research could adopt multi-wave or longitudinal designs to better establish causal inferences among leader behavior, psychological safety, and prohibitive voice. Additionally, researchers could further incorporate procedural remedies—such as using marker variables or collecting data at different time points—to enhance the validity and robustness of findings⁷⁴.

Second, the sample was drawn entirely from Chinese organizations, potentially reflecting cultural influences such as power distance or face-saving. While this context is suitable for studying prohibitive voice, generalizability remains limited. Future studies should replicate or contrast the model in other East Asian societies (e.g., Korea or Japan) or Western countries to explore how national culture moderates leader voice dynamics^27,85.

Third, although this study focused on psychological safety as the core mediating mechanism, future research could explore alternative or complementary psychological pathways that may help explain the underlying process more fully. For instance, affective states such as positive affectivity may also shape whether employees perceive voice behavior as rewarding or risky. Employees who experience frequent positive emotions may be more inclined to engage in voice behavior, regardless of external cues, by virtue of their enhanced psychological resilience and optimism⁸⁶. Incorporating such affective mediators would enrich the understanding of the emotional dynamics underlying employee voice and expand the current framework beyond cognition-focused mechanisms.

Fourth, while the current study considered generational cohort and self-efficacy as important individual-level moderators, future studies should extend the model to include broader contextual moderators such as team characteristics or task-related features. For example, team voice climate—the shared belief among team members that speaking up is safe, encouraged, and valued—may significantly influence how leader cues are interpreted and acted upon⁸⁷. Likewise, task interdependence or task uncertainty could moderate the relationship between leader behavior and employee voice, such that employees in highly interdependent or ambiguous tasks may be more likely to engage in prohibitive voice due to collective exposure to risk and stronger perceived responsibility⁸⁸. By incorporating such situational moderators, future research could contribute to building a more comprehensive, multilevel framework that better captures the complexity of conditions under which prohibitive voice emerges.

Fifth, although the present study focused exclusively on follower prohibitive voice to investigate how leader behavior and contextual signals shape risk-laden voice behavior, this approach inevitably limited the scope of our findings. Given that promotive and prohibitive voice often co-occur and may share overlapping antecedents, future research should adopt a broader, parallel modeling approach that incorporates both leader and follower promotive and prohibitive voice behaviors. Such a 2 × 2 framework would provide a more comprehensive understanding of how different forms of leader voice uniquely and jointly influence distinct follower voice outcomes. We recognize this as an important extension and are currently exploring this expanded model using the same dataset in a follow-up study.

Finally, future research could also explore the applicability of alternative theoretical frameworks beyond UMT, thereby advancing theoretical pluralism in the study of prohibitive voice behavior. While UMT provides a robust lens for understanding how employees use social cues to manage uncertainty, other theoretical perspectives may offer complementary or novel insights into the mechanisms and boundary conditions that shape prohibitive voice. For example, Social Information Processing Theory⁸⁹ may serve as a useful framework for explaining how employees construct meaning from their social environment and interpret leader behaviors in making voice-related decisions. This theory emphasizes that individuals’ attitudes and behaviors are shaped by the information embedded in their immediate social contexts. Applying this perspective could extend current findings by examining how communication patterns within teams, peer reactions, or even organizational rituals influence prohibitive voice, beyond leader behavior alone.

Another promising theoretical lens is Affective Events Theory⁹⁰ which highlights the role of discrete emotional experiences at work in shaping individual behavior. While the current study conceptualizes psychological safety as a cognitively grounded mediator, future research could examine how affective events—such as witnessing organizational misconduct or receiving harsh feedback—elicit emotional reactions such as anger, anxiety, or moral outrage that may subsequently drive prohibitive voice. This approach would deepen our understanding of the emotional antecedents of voice and identify distinct affective pathways that may operate independently from perceptions of psychological safety or self-efficacy.

By integrating these alternative theoretical perspectives, future research can move beyond a purely uncertainty-management lens to develop a more holistic and multidimensional understanding of prohibitive voice. Such theoretical expansions would not only enrich conceptual clarity but also offer more tailored and effective strategies for promoting voice across diverse organizational contexts and employee profiles.

Conclusion

This study advances the understanding of prohibitive voice behavior by developing and testing a multi-dimensional framework grounded in Uncertainty Management Theory. By examining leader prohibitive voice as a behavioral cue, psychological safety as a mediating mechanism, and self-efficacy and generational cohort as boundary conditions, the study provides a comprehensive account of how employees navigate uncertainty in deciding whether to speak up about potential risks or problems. The findings underscore the powerful role of leader behavior in shaping employee voice, while also revealing important variations in how such cues are interpreted depending on individual and generational differences. Moreover, the results highlight the centrality of psychological safety in enabling upward dissent, particularly for employees with lower self-efficacy or those from younger generations. Collectively, these insights not only enrich the theoretical landscape of voice behavior by extending the scope of UMT, but also offer practical guidance for cultivating leadership practices and organizational climates that support constructive voice. In doing so, this research contributes to ongoing efforts to build more open, responsive, and ethically resilient organizations.