An employee notices that a scheduling rule creates avoidable errors. She mentions it to a colleague, writes a private note, and says nothing in the next meeting. Later, she submits a concrete proposal through a reporting channel. Which moment counts as voice? The answer depends on whether the study concerns expressed ideas, the intended recipient, the opportunity to influence decisions, or the choice to withhold a concern. Merely counting spoken words will miss the behavior the study wants to understand.
Measuring voice behavior means defining discretionary communication of work-relevant ideas, concerns, or suggestions and identifying when that communication actually occurred. Voice can promote improvement or warn against harm. It can move upward, sideways, or through a formal system. This guide compares self-report, supervisor ratings, direct observation, and event-based records. It keeps voice distinct from psychological safety, silence, agreement, and eventual organizational change, which may relate to voice but are not interchangeable measurements.
Set the boundary of voice
Van Dyne and LePine examined helping and voice as distinct extra-role behaviors and reported construct and predictive validity in a field study. Their approach made voice more specific than general helpfulness or job performance. A measurement item should identify expression of a constructive idea or concern intended to improve work. Polite conversation, routine status reporting, and compliance with an assigned speaking duty may have different meanings. The boundary depends on the study’s question and should be written before coding begins.
Voice is not always spoken aloud. An employee can raise a concern in a written incident report, message, or design review. Conversely, loud participation in a meeting can involve little new information. Define the channel and intended audience. If the research asks about upward voice to a manager, a private discussion with a peer is a different behavior. If it asks about circulation of improvement ideas across a team, that peer discussion may be central. Neither definition is automatically superior.
Liang and colleagues distinguished promotive voice, which offers ideas for improvement, from prohibitive voice, which warns about harmful practices or risks. Their two-wave study measured these forms separately. This distinction matters because the interpersonal cost of saying “we could do better” can differ from saying “this plan is unsafe.” A single positive-voice score may conceal whether a team welcomes suggestions while punishing warnings. Preserve content categories when the question concerns risk reporting or prevention.
Voice is also an act, not proof of impact. An employee may submit a strong suggestion that is ignored. Another may make a weak suggestion that is adopted for unrelated reasons. Measure expression and uptake separately. A successful change can follow many inputs, while a voiced concern may be valuable even if no immediate change follows. Treating implementation as part of the voice score would make it hard to study whether organizations respond fairly to ideas.
Ask people about their own behavior
Self-report can capture voice that a supervisor never heard or does not remember, especially across informal channels. It also lets a participant identify the issue and intended audience. But retrospective items such as “I often speak up” depend on memory and self-image. A person may recall a courageous occasion while overlooking repeated routine silence. Specify a time window and ask about concrete episodes or categories rather than relying solely on a broad trait-like summary.
Frequency needs a denominator. Five suggestions in a month may represent frequent voice in a quiet role or rare voice in a role with daily opportunities to identify problems. Ask whether the participant had ideas or concerns worth raising, whether a relevant recipient was available, and whether the issue was within their knowledge. A low count without opportunities is not the same as withholding. Morrison’s later review stresses that many voice measures do not allow low voice to be interpreted automatically as silence because a person may have had nothing to say.
Confidentiality is part of validity. If an employer can identify who reported criticism, respondents may understate voice or overstate it to look proactive, depending on incentives. Explain who receives raw data and how small-group results are protected. A survey about warnings of safety failures should not be administered in a way that exposes the few people who raised those warnings. Ethical handling improves the chance that a score describes actual behavior.
Self-report also varies in what respondents call “speaking up.” One person may count a mild question, another only a direct challenge. Provide examples and distinguish a new proposal from ordinary agreement. If comparing roles or cultures, pilot the wording to see whether people interpret “voice,” “concern,” and “suggestion” similarly. A literal translation may not preserve the social meaning of challenging a senior colleague.
Use other ratings without making managers the sole witness
Supervisor ratings can assess behaviors visible to managers, and they can be linked to other organizational records. Detert and Burris studied leadership behavior and employee voice using a measure focused on verbal behavior. Their design shows a common approach, but a manager rating is not a complete record. A supervisor may miss private reports, discount a critical employee, or remember only dramatic incidents. State whose perception the score represents and which channels that rater could observe.
Peer ratings can add another view, especially when colleagues hear suggestions before managers do. Yet peers differ in exposure and relationship to the speaker. If several raters are used, report agreement and examine whether disagreement is systematic. A person might be vocal in technical reviews and quiet in executive meetings. Averaging those reports can hide a context-specific pattern that is more informative than the mean.
A leader’s openness can influence whether employees speak, but it can also influence how the leader scores their voice. If the same manager rates both their own openness and subordinates’ voice, a positive association may partly reflect a common perspective. Use separate sources where possible and distinguish an antecedent from the behavior. Detert and Burris’s findings do not mean that a cheerful invitation to speak guarantees that every issue reaches the leader.
Group voice climate is another distinct variable. Morrison, Wheeler-Smith, and Kamdar examined how shared group beliefs about voice relate to individual voice behavior. A climate scale measures expectations in a group; an individual voice measure records what a person did or was perceived to do. Aggregation at the group level requires evidence of shared belief, while behavior can vary among members. Keeping levels separate prevents a group reputation from being mistaken for each employee’s action.
Observe episodes and channels
Direct observation can capture who raised a concern, what it concerned, when it occurred, and how recipients responded. A coding scheme should distinguish a question seeking information, a suggestion for improvement, a warning about risk, and a complaint without a proposed remedy. Not every criticism is unhelpful; a warning can be valuable without an immediate solution. The coding rule should be content based and applicable by independent observers rather than determined by whether a leader liked the message.
Meeting observations sample only public moments. Employees may choose a private channel because the issue is sensitive or because a formal report is required. If the aim is organization-wide voice, combine meeting coding with appropriately protected records of reports, suggestions, and follow-up. Administrative counts can be misleading if a new portal makes submission easier or if policy changes require formal documentation. A rising count may reflect access to a channel rather than a change in willingness.
Event-based diaries can reduce long recall periods by asking participants soon after a potential issue arose whether they shared it, with whom, and why. They also create burden and may make people monitor themselves differently. Define the entry trigger carefully. Asking only after a voiced episode will never capture withholding. Asking about every minor dissatisfaction may flood the data with situations that were not meaningful opportunities for voice. A feasible diary asks about a work-relevant idea or concern and then tracks expression and nonexpression.
The uptake of voice should be coded separately: acknowledgment, request for detail, investigation, dismissal, retaliation, or later action. These outcomes can influence future voice but do not retroactively change whether the original act occurred. A team may invite ideas yet ignore them, or respond well only to promotive suggestions. Separating expression from response makes those patterns visible. It also helps explain why a measure of voice alone cannot certify a psychologically safe culture.
Do not infer silence from zero voice
Silence concerns information, ideas, or concerns a person could have shared but withheld. A person who has no relevant issue to raise may be quiet without engaging in silence in that sense. Sherf, Parke, and Isaakyan examined voice and silence as distinct behaviors with different relationships to perceived impact, psychological safety, and burnout. Their study supports measuring both rather than deriving one mechanically as the absence of the other.
The difference creates a difficult observation problem: a researcher can hear speech but cannot directly observe an unspoken idea. Private diaries, interviews, and confidential issue-specific surveys can provide evidence about withholding. They remain self-reports subject to recall and disclosure limits. A meeting video that shows no challenge establishes only that none was voiced in that recorded period. It does not prove that participants had concerns and suppressed them.
Strategic timing also matters. Someone may wait until they have evidence, choose a private audience, or avoid interrupting a crisis. Not every delay is fear-based silence. Ask what was known, when, to whom it could have been communicated, and whether it was eventually raised. A study that labels all waiting as suppression risks punishing thoughtful judgment. Conversely, a later report should not erase the cost of an earlier missed opportunity if the delay mattered.
High-stakes topics can be especially undercounted. Harassment, discrimination, unsafe practice, or fraud may involve legal and personal consequences. A generic survey item about suggesting improvements may poorly represent these issues. Researchers should avoid asking for identifiable details in an insecure channel. If the study includes severe concerns, design confidentiality and referral procedures before data collection. The quality of the measurement depends on whether people can answer without additional exposure.
Interpret content, opportunity, and power
Promotive and prohibitive voice may have different base rates and reception. Liang and colleagues’ separation of forms is useful when an organization responds warmly to innovation ideas but defensively to warnings. Report each form rather than averaging them into a positivity score. The value of an issue is not determined by its tone: an uncomfortable warning can be the most important contribution in a high-risk setting.
Power shapes both the cost and visibility of voice. Senior people may speak more often because the floor is theirs, while junior people may submit ideas privately or remain silent. A raw count of voiced proposals may reward authority rather than willingness to challenge. Record role, opportunity to speak, channel, and audience. Compare like with like where possible, and avoid treating a lower count among less powerful members as a personal deficit.
Repetition can mean persistence or failure of uptake. If the same safety concern is raised five times because no one acts, five voice events do not indicate five independent improvements in climate. Code issue identity and the sequence of responses. A system that counts submissions without resolving them may look lively while exhausting contributors. Voice behavior is useful to measure, but organizational response determines much of its consequence.
Organizational changes can shift where voice appears. A new anonymous channel may increase formal reports while public meeting voice declines. This might represent better access, reduced trust in managers, or both. Evaluate channels together and ask participants why they chose them. A single-platform count cannot support a broad claim about employees’ willingness to speak. The intended audience and expected impact are part of the construct.
What the evidence does not show
The cited studies do not show that more voice is always better, that a quiet person lacks ideas, or that every voiced suggestion improves decisions. Voice scores depend on opportunities, content, channel, audience, and rater. A supervisor rating can reflect visibility as well as behavior. An administrative count can reflect policy as well as participation. A low score cannot be translated into silence without evidence that a relevant issue existed and was withheld.
Nor does voice alone establish psychological safety. People sometimes speak despite fear because a risk is urgent, a role requires it, or they expect an external audience to listen. Others may feel safe but have no reason to raise a concern during the sampled period. The Institute’s synthesis is to maintain an explicit chain: issue recognized, opportunity present, expression or withholding chosen, recipient response, and later consequence. Each link calls for its own evidence.
Make the result interpretable
A defensible report defines promotive and prohibitive content, the sampled channels, the observation window, who rated whom, response rates, and how opportunities were estimated. It explains whether the score is event frequency, a survey scale, or a recipient’s judgment. It reports voice and silence separately if both are claimed. When records include sensitive issues, it protects identities and does not publish small-cell patterns that expose contributors.
In the opening example, the employee’s later proposal is a voice event. The earlier private note is evidence of a recognized issue, not necessarily communication to an actionable audience. Her quiet meeting may or may not have been silence, depending on whether she deliberately withheld the issue there. Measurement becomes useful when it can tell those moments apart and show what happened after the idea finally reached someone able to act.
Research references
- Van Dyne, L., & LePine, J. A. (1998). Helping and voice extra-role behaviors: Evidence of construct and predictive validity. Academy of Management Journal, 41, 108–119. View source
- Liang, J., Farh, C. I. C., & Farh, J.-L. (2012). Psychological antecedents of promotive and prohibitive voice: A two-wave examination. Academy of Management Journal, 55, 71–92. View source
- Detert, J. R., & Burris, E. R. (2007). Leadership behavior and employee voice: Is the door really open? Academy of Management Journal, 50, 869–884. View source
- Morrison, E. W., Wheeler-Smith, S. L., & Kamdar, D. (2011). Speaking up in groups: A cross-level study of group voice climate and voice. Journal of Applied Psychology, 96, 183–191. View source
- Sherf, E. N., Parke, M. R., & Isaakyan, S. (2021). Distinguishing voice and silence at work: Unique relationships with perceived impact, psychological safety, and burnout. Academy of Management Journal, 64, 114–148. View source
- Morrison, E. W. (2023). Employee voice and silence: Taking stock a decade later. Annual Review of Organizational Psychology and Organizational Behavior, 10, 79–107. View source