When Speaking Up Divides: Voice Modality Divergence, Behavioral Faultlines, and Team Performance

1. Introduction

Speaking up is one of the most central interpersonal behaviors through which teams learn, adapt, and improve (Edmondson, 1999; Morrison, 2011). Through voice, employees surface new ideas, highlight emerging problems, and shape more informed collective decisions. As a result, existing research has developed a robust understanding of voice antecedents and outcomes—leadership (Detert & Burris, 2007), psychological safety (Edmondson & Lei, 2014), and individual dispositions (LePine & Van Dyne, 1998), among others—creating a strong consensus that voice is generally beneficial for teams.

Yet an essential question remains underdeveloped: How do different forms of voice combine within teams, and what team-level structures do these combinations produce? Voice is not a monolithic behavior. Promotive voice focuses on articulating ideas for improvement and innovation, whereas prohibitive voice raises concerns intended to prevent failures or mitigate harm (Liang, Farh, & Farh, 2012). These modalities differ in tone, purpose, and motivational underpinnings, but research has rarely conceptualized them as a patterned team-level system.

To address this gap, we introduce Voice Modality Divergence (VMD) as a new team-level composition construct capturing the extent to which a team collectively differentiates and balances promotive and prohibitive voice. Drawing on theories of ambidexterity (Gibson & Birkinshaw, 2004), information integration (van Ginkel & van Knippenberg, 2009), and team learning (Edmondson, 1999), we propose that VMD enhances team performance by enabling teams to integrate improvement-oriented ideas with prevention-oriented concerns. In complex environments, teams must simultaneously pursue opportunities and manage risks (March, 1991; O’Reilly & Tushman, 2013). VMD provides a behavioral foundation for enacting this dual imperative.

At the same time, we argue that voice configurations can generate not only informational resources but also social structures that shape whether teams function cohesively. Building on faultline theory (Lau & Murnighan, 1998; Thatcher, Jehn, & Zanutto, 2003), we introduce Voice-Based Faultline Strength (VFS)—a behavioral segmentation that arises when voice behaviors align such that high voice (both promotive and prohibitive) concentrates within a subset of members while low voice concentrates within another. Such alignment can sharpen subgroup boundaries, trigger social categorization processes, and fragment communication and psychological safety (Hogg & Terry, 2000; Meyer et al., 2015). Importantly, in this revised model, we theorize VFS not as a moderator but as a direct performance liability: teams suffer when voice becomes structurally segregated, regardless of whether the team exhibits balanced voice overall.

By theorizing VMD as a team resource and VFS as a team structural constraint, this study advances voice theory beyond individual-level predictors to the collective architecture through which speaking up is distributed. We propose and test two direct hypotheses linking VMD and VFS to team performance. In doing so, we offer a more nuanced explanation for why speaking up sometimes helps teams excel and other times divides them.

2. Theoretical Background and Hypotheses

2.1. Voice Behavior as a Complex, Multi-Modal Team Resource

Voice behavior has long been recognized as a critical driver of team learning and effectiveness (Edmondson, 1999; Morrison, 2011). However, a growing body of work acknowledges that voice comprises distinct modalities serving different psychological and functional purposes (Liang et al., 2012; Morrison, 2014). Promotive voice facilitates exploration, ideation, and change-oriented thinking, whereas prohibitive voice supports vigilance, risk detection, and harm prevention (Van Dyne & LePine, 1998; Li & Sun, 2015). In complex work systems, both are valuable, yet they can be difficult to combine because they reflect different cognitive frames (opportunity vs. risk) and different interactional tones (aspirational vs. cautionary).

Despite this, much voice research continues to treat promotive and prohibitive voice as independent predictors, typically aggregating voice frequency or intensity without capturing how modalities are configured and balanced at the team level (e.g., Fast et al., 2014). As Ashford et al. (2020) note, voice research remains insufficiently sensitive to within-team structures that govern how voiced inputs are integrated.

2.2. Voice Modality Divergence and Team Performance

We propose Voice Modality Divergence (VMD) as a team-level composition construct capturing the degree to which a team collectively differentiates and balances promotive and prohibitive voice. The theoretical logic rests on three complementary streams.

First, research on information integration and elaboration suggests that teams perform better when they incorporate divergent perspectives and engage in deeper processing of heterogeneous inputs (van Ginkel & van Knippenberg, 2009; Kearney, Gebert, & Voelpel, 2009). Promotive voice supplies improvement-oriented ideas and alternatives; prohibitive voice surfaces risks, constraints, and latent failures. When teams exhibit high VMD, they are more likely to generate comprehensive solutions that incorporate both upside opportunities and downside protections.

Second, VMD reflects a behavioral foundation for ambidexterity. Ambidexterity research emphasizes balancing exploration and exploitation (March, 1991; Gibson & Birkinshaw, 2004) and integrating paradoxical demands (Smith & Lewis, 2011). Promotive and prohibitive voice map onto these dual demands: innovation-oriented exploration and prevention-oriented reliability. High VMD indicates that both modalities are present in meaningful degrees, enabling teams to pursue improvement without neglecting risk control.

Third, team learning theories suggest that promotive voice stimulates experimentation and adaptation while prohibitive voice supports error detection and correction (Edmondson, 1999; Tucker & Edmondson, 2003). Integrating both modalities should strengthen learning cycles and resilience, resulting in higher performance.

Hypothesis 1.

Voice Modality Divergence (VMD) is positively associated with team performance.

2.3. Voice-Based Faultline Strength as a Divisive Behavioral Structure

While VMD captures the balance of modalities, teams may simultaneously develop behavioral subgroup structures that hinder coordination. Faultline theory explains that when attributes align, teams split into homogeneous subgroups, which intensifies social categorization, reduces communication, and impairs coordination (Lau & Murnighan, 1998; Thatcher et al., 2003). Although faultlines were originally studied using demographic attributes, recent work emphasizes deep-level and behavior-based attributes as potent bases for subgroup formation (Meyer & Glenz, 2013; Carton & Cummings, 2013). Behavioral configurations are especially consequential because they directly govern interaction patterns and interpretive processes.

We define Voice-Based Faultline Strength (VFS) as a behavioral segmentation that emerges when promotive and prohibitive voice align such that high voice of both modalities concentrates within one subgroup, while low voice of both concentrates within another. This configuration creates rigid subgroups separated by differential participation, influence, and visibility in team discussions.

We theorize that VFS directly undermines team performance for three reasons.

First, communication becomes structurally restricted. High-voice members exchange more information within their subgroup, while low-voice members remain peripheral. Cross-subgroup elaboration declines, leading to incomplete integration of task-relevant information (Carton & Cummings, 2013; Meyer et al., 2015). Second, misattributions and conflict intensify. High-voice subgroup members may interpret low-voice members as disengaged or incompetent, whereas low-voice members may interpret high-voice members as domineering or critical. Such attributional divergence increases task conflict and can spill over into relational conflict (Jehn, 1995). Third, psychological safety becomes fragmented. Subgroups develop localized climates of safety and participation; members may feel safe within their subgroup but not across subgroup boundaries. This fragmentation reduces the likelihood that concerns are voiced and heard across the entire team, lowering coordination quality and collective effectiveness (Meyer et al., 2015; Edmondson & Lei, 2014).

Thus, even if teams have valuable voiced inputs, strong VFS creates a social architecture that impairs collective functioning and reduces performance.

Hypothesis 2

.Voice-Based Faultline Strength (VFS) is negatively associated with team performance

3. Method

3.1. Sample and Data Collection

Data were collected from intact work teams drawn from knowledge-intensive firms. Individual-level voice behaviors and team performance were assessed via team leaders. Of the 47 teams, 41 were retained in the final sample after excluding teams with insufficient member responses to ensure reliable aggregation. The final sample had an average team size of approximately 6 members. Teams operated in functions such as R&D, operations, and consulting, and average team tenure was approximately 19 months. Prior to aggregation, we examined within-team agreement (rwg) and between-team variance (ICC[1]) to justify team-level analyses.

3.2. Variables and Measures

Promotive and Prohibitive Voice Promotive voice and prohibitive voice were measured using the scale developed by Liang, Farh, and Farh (2012). Promotive voice was assessed with 5 items (e.g., “I proactively suggest new ideas to improve work processes”), and prohibitive voice with 5 items (e.g., “I speak up to prevent potential problems or losses”), using a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). Cronbach’s alphas were α = .80 for promotive voice and α = .89 for prohibitive voice.

Voice Modality Divergence (VMD) VMD captures the extent to which a team collectively balances promotive and prohibitive voice. Following prior work on team composition and configurational measures, we computed an individual-level modality balance index as the absolute difference between promotive and prohibitive voice. Higher values indicate greater modality imbalance at the individual level. We then aggregated this imbalance to the team level and reverse-coded the index so that higher values reflect greater balance and divergence in the sense of meaningful presence of both modalities. Thus, higher VMD indicates teams whose members, on average, exhibit more balanced patterns of promotive and prohibitive voice rather than reliance on a single modality.

Voice-Based Faultline Strength (VFS) VFS was computed using Thatcher, Jehn, and Zanutto’s (2003) faultline algorithm, adapted to behavioral data. Individual promotive and prohibitive voice scores were entered as faultline attributes. Higher VFS indicates stronger alignment of voice behaviors across members, producing homogeneous subgroups (high-high vs. low-low) and greater subgroup segmentation. Higher values of VFS indicate stronger alignment of voice modalities across members, such that individuals high in both promotive and prohibitive voice cluster together while individuals low in both modalities form a separate subgroup. This operationalization captures the extent to which voice behaviors are structurally segmented into subgroups, consistent with our conceptualization of behavioral faultlines.

Team Performance Team performance was assessed using leader-rated measures (e.g., “This team meets or exceeds its performance goals”), aggregated to a team-level score.

Control Variables We controlled for team size, team tenure, and demographic diversity (e.g., age or gender diversity). We also controlled for average promotive voice and average prohibitive voice to separate configurational effects from overall voice level.

4. Results

4.1. Descriptive Statistics and Correlations

As shown in the Table 1, promotive voice and prohibitive voice were moderately correlated, indicating that although related, they represent distinct constructs. Voice Modality Divergence was positively correlated with team performance (reverse), providing preliminary support for Hypothesis 1. Voice-Based Faultline Strength was negatively correlated with team performance and with VMD provided that introductory support for Hypothesis 2.

Intraclass correlation coefficients (ICC[1]) and within-group agreement indices (rwg) supported aggregation of individual-level voice measures to the team level (ICC[1]s > .10; rwg > .70).

4.2. Hypothesis Testing

Hypothesis 1 predicted that Voice Modality Divergence would be positively related to team performance. As shown in Model 3 of Table 2, VMD (reverse) was positively and significantly associated with team performance (b = -.349, p < *), controlling for team characteristics and average levels of promotive and prohibitive voice. Thus, Hypothesis 1 was supported. Hypothesis 2 predicted that Voice-Based Faultline Strength would be negatively related to team performance. Results in Model 3 show that VFS was negatively and significantly associated with team performance (b = -.293, p < *). This finding indicates that teams whose voice behaviors are segmented into homogeneous subgroups perform worse, net of overall voice levels. Thus, Hypothesis 2 was supported. Robustness checks using alternative operationalizations of VMD and alternative performance indicators yielded substantively consistent results.

5. Discussion and Conclusions

This study advances the understanding of voice in teams by shifting attention from the frequency of speaking up to the structural configuration of voice modalities and the social segmentation that emerges around them. Our results demonstrate two central insights.

First, teams benefit when promotive and prohibitive voice are collectively balanced and differentiated. Voice Modality Divergence (VMD) captures this configurational resource and is positively associated with team performance. This finding suggests that teams perform better when they combine improvement-oriented and prevention-oriented inputs, enabling more comprehensive decision-making, stronger learning cycles, and greater resilience. Second, teams suffer when voice behavior becomes structurally segregated into homogeneous subgroups. Voice-Based Faultline Strength (VFS) directly undermines team performance, above and beyond average levels of voice. This supports the view that voice is not only informational but also relational and structural. When high voice concentrates within a subset of members while others remain systematically low-voice, teams become fragmented in communication, develop divergent attributions, and experience fractured psychological safety. In such cases, teams lose the integrative capacity needed for effective coordination and collective learning.

5.1. Theoretical Contributions

This study contributes to voice research by conceptualizing voice as a team-level configurational phenomenon rather than a simple aggregate of individual behavior. VMD captures how teams collectively balance two distinct modalities, extending voice research toward compositional and structural accounts of team processes. Second, we bridge voice and ambidexterity by showing that promotive and prohibitive voice operate as complementary behavioral inputs supporting exploration and reliability simultaneously. VMD provides behavioral microfoundations for how teams enact ambidexterity in everyday interaction. Third, we extend faultline theory by conceptualizing behavioral faultlines around enacted voice behaviors. By introducing Voice-Based Faultline Strength (VFS), we demonstrate that subgroup divisions can emerge not only from who members are but from how they participate in collective communication. Finally, our direct-effects framework clarifies that voice diversity has dual consequences: configurational balance enhances performance, while behavioral segmentation undermines it. This offers a more conditional and structurally grounded understanding of when speaking up helps teams and when it divides them.

5.2. Practical Implications

Leaders should cultivate environments that encourage both promotive and prohibitive voice, legitimizing the value of improvement and prevention inputs. At the same time, leaders must monitor whether voice becomes concentrated within a subset of members. Teams may appear “high-voice” overall while still suffering performance losses if voice participation is segregated. Practices such as structured turn-taking, rotating facilitation, and deliberate cross-member problem-solving can reduce voice-based segmentation and support collective integration.

5.3. Limitations and Future Research

This study relies on a relatively small number of teams, which may limit generalizability. Future work should replicate findings across industries and cultures and examine longitudinal dynamics of how VMD and VFS evolve over time. Research could also test mediators (e.g., information elaboration, attribution patterns, subgroup psychological safety) to directly capture mechanisms through which VMD enhances performance and VFS undermines it.

6. Conclusions

Speaking up is central to team learning and performance, but its benefits depend on how voice behaviors are configured and socially structured. Teams perform better when promotive and prohibitive voice are collectively balanced, yet they perform worse when voice patterns become segregated into subgroups. By introducing VMD and VFS as distinct team-level constructs with independent performance implications, this study advances a structurally grounded account of how voice can both unite and divide teams.