Voice, Silence and Speaking Up
How voice, silence, power, interpersonal risk and the response to speaking up interact, and where the research disagrees.
The landscape
Voice, as the voice literature uses the word, is informal, discretionary, improvement-oriented communication of ideas, concerns or problems upward to someone who could act on them, and silence is the withholding of such input rather than the mere absence of speech: someone with nothing to say is not silent in this sense (Morrison (2014)). Psychological safety came into the picture as a shared belief that a team is safe for interpersonal risk taking, which Edmondson (1999) was careful to say is not cohesiveness and not permissiveness, involves but exceeds trust, and is a confidence that the team will not embarrass, reject or punish someone for speaking up.
Our own way in is the calculus of voice: before every act of voice a rapid, usually unconscious risk assessment runs, weighing the potential costs of speaking (lost status, exclusion, embarrassment, punishment, more work) against the potential benefits, and it produces four outputs rather than two: speaking fully, silence, softening, and changing the message to please or appease. Tom's piece also widens voice to any intentional act of communication, not only speech, because narrowing it to speech excludes exactly those who most need to be heard.
Four questions recur. Why people stay silent when they have something to say, when Milliken, Morrison & Hewlin (2003) found the most common topics withheld were concerns about management behaviour and organisational direction. Whether safety and effectiveness are one judgement or two (Sherf, Parke & Isaakyan (2021)). What happens after someone speaks (How We Respond). And who voice is available to at all, when Kanter (1977) explains why the same team can be safe for the many and unsafe for the few.
The key connections
The decision to speak. Dutton & Ashford (1993) are the historical root: issue selling, in which people raise an issue when they both value it enough and expect the attempt to succeed. Detert & Burris (2007) named the affect-laden expectancy calculus that employees run before speaking up, and Morrison (2014) set two judgements at the centre of the field: efficacy, whether voice will change anything, and safety, whether it will rebound on the speaker. Tom's two-gate framework is the same pair, ambiguity failure (is it safe to speak?) and valence failure (will it matter?), and Detert & Treviño (2010) found which gate matters where: for immediate supervisors futility outweighs fear by roughly 1.8 to 1, while for leaders two to five levels up the balance tips towards lack of safety. Our own barriers survey found 'I don't think it will make a difference anyway' the single most common barrier to speaking up.
Silence as its own thing. Morrison & Milliken (2000) introduced organisational silence as a collective phenomenon, the widespread withholding of concerns, information and opinions, and Milliken, Morrison & Hewlin (2003) supplied the employee side. Pinder & Harlos (2001) split silence in two: quiescence, the conscious, uncomfortable withholding of someone who has not given up, and acquiescence, the deeper resignation of someone who accepts the circumstances as normal, and the route out of the second runs through the first. Van Dyne, Ang & Botero (2003) sorted both silence and voice by motive, acquiescent, defensive and prosocial, and Kish-Gephart, Detert, Treviño & Edmondson (2009) crossed fear intensity with the time available to respond and got four kinds of silence, from an automatic freeze to the deliberate weighing the literature had treated as the whole story. Tom's Types of Silence draws on the five types in Pacheco et al (2015) and adds two of ours, preparatory and filled.
After someone speaks. Jade's How We Respond puts the onus on the listener: you can tell people to speak up endlessly, but if your past, present or anticipated responses involve punishing, blaming or humiliating them, they won't. The observer effect is the same mechanism at one remove: people watch how leaders respond to good and bad information and use it to predict how safe it is for them to take a similar risk. Noort, Reader & Gillespie (2021) demonstrated the second half at scale, coding 14,128 conversational turns from 172 cockpit voice recorder transcripts: safety voice occurred in all but two of the accidents, and what varied was the response, which they call safety listening. Kerrissey, Satterstrom, Pae & Albert (2024) found that voiced ideas hit walls or fall into voids, and Edmondson (2003b) has the most vivid illustration of a good response: a surgeon who said nothing when a nurse dropped a freshly harvested vein graft and simply went back and harvested another.
Who gets to speak. Goffman (1955) is the source Edmondson reaches for on the cost of speaking: by saying something, a speaker opens the possibility of being thought forward, foolish or offensive, and there is much to be gained from venturing nothing. Bourdieu (1991) argues that whose speech counts as legitimate is set by the social position of the speaker, and that speaking up needs not just permission but a felt entitlement to speak, which is accumulated rather than given by climate. Kanter (1977) explains why the same team can be safe for the many and unsafe for the few: tokens, at roughly fifteen per cent or less of a group, are watched, stereotyped and structurally isolated. Brescoll (2011) and McClean et al (2018) show the penalty, and the missing reward, for women who speak up; Milanovich et al (1998) show cockpit rank being read as competence on every measure, flying-related or not; and Weick (1990) shows the steepening authority gradient at Tenerife making the junior crew's doubts progressively harder to voice. Crenshaw (1989) supplies the reason measuring safety for each group separately will not add up to the truth, and our Least Safe Person argues that a team average hides the outlier who is suffering.
The tensions and disagreements
Is silence a calculation at all? Kish-Gephart et al (2009) argue that a deliberate calculus describes only one corner of the picture, because a great deal of silence never goes through anything resembling deliberation, and Morrison (2014) agrees that much silence is automatic and emotionally driven, running on the implicit voice theories that Detert & Edmondson (2011) found operating below awareness. Our calculus of voice pushes back: a computation does not stop being a computation for running too fast to notice, and it reads their four types as four speeds and routes the calculus can take.
Are voice and silence opposites? For years the field assumed that whatever raises voice lowers silence. Van Dyne, Ang & Botero (2003) argued otherwise on theoretical grounds, and Sherf, Parke & Isaakyan (2021) showed it: across 162 papers the corrected correlation between voice and silence was weak (r = -.15); perceived impact predicts voice more strongly than silence, and psychological safety predicts silence more strongly than voice. The consequence, in Morrison's (2023) reading, is that research cannot simply flip the sign on voice predictors to explain silence; the practical consequence, in Sherf, Parke & Isaakyan's own, is that building safety alone may not be enough to increase voice. Our barriers survey found the same shape in the field: a Cynical-Conformist archetype that is not scared, just not convinced it matters.
Is silence always bad? Pinder & Harlos (2001) say silence is not uniformly bad and voice is not uniformly good: for some people acquiescence may be the most adaptive response available, and a group in constant complaint is dysfunctional too. Van Dyne, Ang & Botero (2003) name prosocial silence, staying quiet to protect a colleague, and Edmondson & Besieux (2021) name processing, the productive silence of listening. Burris, Detert & Chiaburu (2008) add that silence sometimes signals that people have already mentally left, not that they feel unsafe. Tom's Power of Silence reaches the same place from the meeting room: silence is neither uniformly a sign of low psychological safety nor of high.
Do formal channels help? Detert & Treviño (2010) found that the more leaders lean on open-door policies, hotlines and ombudspersons, the less safe employees may feel, because the machinery concedes that speaking up is risky ('Who pays the ombudsman's salary?'). Morrison (2023) reports employees with more formal channels using them for helpful ideas while becoming more selective about anything that might disturb group harmony, and our barriers survey heard doubts that anonymous channels are truly anonymous. Against that, Sherf, Parke & Isaakyan (2021) suggest safety-generating mechanisms such as anonymous channels or protection from punishment to reduce silence. This one is unresolved.
Trust and psychological safety. Edmondson (1999) could not conclusively separate the construct from trust, and said so. Tom's PS & Trust argues the two are related but not interchangeable: high cognitive trust can work against challenge, because the more we trust someone's expertise the less we credit our own dissenting read, and reducing psychological safety to trust locates the problem in the warmth of individual relationships rather than in structural incentives and power gradients.
What counts as voice. Morrison (2023) defines voice narrowly and deliberately, excluding peer-only voice and external whistleblowing, and reports the employment relations charge that the organisational behaviour literature is a narrow, managerialist silo that ignores collective and formal mechanisms such as unions and works councils; Wilkinson et al (2018) make that charge from the other side, and turn it on their own field for narrowing voice to the airing of grievances. Kassing (1998) anchors a dissent tradition that ran parallel to the voice research for two decades with little traffic between them, asking where disagreement goes when it cannot go up. Our calculus of voice widens the definition on a different axis, to any intentional act of communication rather than speech alone, and Jade's Whistleblowing & PS places whistleblowing at the far end of a spectrum of disclosure: what happens when routine voice has failed.
Blame and no blame. Dekker (2011) shows criminal prosecution after professional mistakes suppressing the voluntary error reporting through which safety cultures learn. Sherratt et al (2023) found the opposite failure: dogmatic no-blame redirects blame onto paperwork and procedures, until investigation returns to something like acts of God that nobody can learn from. Tom's Accountability sits between them: accountability as the capacity to give an honest account of what happened, colonised by the imposition of consequences on whoever is nearest to a bad outcome.
What this means for practice
- Two problems, not one. Sherf, Parke & Isaakyan (2021) suggest organisations need impact-generating mechanisms, such as visible follow-through on suggestions, to cultivate voice, and safety-generating mechanisms to reduce silence, rather than assuming either produces the other. Our barriers survey maps three archetypes to different priority interventions and concludes there is no single fix.
- Work on the response. Noort, Reader & Gillespie (2021) argue the field should move from making safety voice more likely to making it more effective, and Jade's How We Respond makes it personal: our reactions when tired, distracted or overwhelmed matter as much as our good-day responses. Kerrissey et al (2024) locate walls and voids in organisational structure and then prescribe largely interpersonal technique, which we read as the failing of a great deal of psychological safety training.
- Mind the gradient, not the junior. Milanovich et al (1998) argue that training junior crew to be more assertive has a limited and sometimes counterproductive payoff, because assertiveness from a low-status speaker can read as a violation of the relationship rather than as information. Edmondson (2003b) and Nembhard & Edmondson (2006) put the lever with leaders: downplaying power differences, inclusiveness towards lower-status members, under-reacting to error. Bourdieu (1991) is the caution: if people have been shaped by years of social experience to experience themselves as not-speakers, no amount of leader openness fully addresses that.
- Measure what is withheld, and the least safe. Sherf, Parke & Isaakyan (2021) found that silence predicts burnout far more strongly than voice does, so tracking how much people say misses the more consequential half; Tom's Least Safe Person argues that a mean score masks the person who is suffering. Edmondson & Kerrissey (2025) add that psychological safety cannot be mandated by policy and varies team to team within a single culture, so it is built interaction by interaction.
- Hold the evidence lightly. Edmondson (1999) is cross-sectional and single-company, and declined to overclaim; Noort, Reader & Gillespie (2021) studied accidents and only accidents, with no normal flights to compare; Hirschman (1970) argued by example about customers, members and citizens, and reading it onto employees is our application.
What the map does not yet hold
Two gaps we know about. Collective and formal voice: Morrison (2023) partly concedes the employment relations critique, and Wilkinson et al (2018) is the one node here that makes it at length; unions and works councils otherwise appear in passing in this region, and Tom's Employment Rights newsletter, which treats unionisation and Works Councils as worker protections that appear to correlate with psychological safety, sits outside it. Ethical voice: Morrison names raising misconduct, harassment or fraud as the field's most significant blind spot, and the map reaches it through Jade's Whistleblowing & PS, Carson (1962), Christopher (2026) on grievance and harassment policies routed through the people they concern, and Siad & Rabi (2021) on harassment as a barrier to safety, rather than through a study of ethical voice as such.
Explore the density
Start here
- Psychological Safety and Learning Behavior in Work Teams
The paper that turned psychological safety from a concept in the organisational change literature into a team-level construct with a measure, and the source of the definition almost every later study…
- Employee Voice and Silence
Morrison's agenda-setting review is the piece that consolidated the scattered voice-and-silence literature into a single field and set its research programme; it is the review her own 2023 'decade la…
- Calculus of Voice
The mechanism beneath psychological safety: before every act of voice a rapid, usually unconscious risk assessment runs — what Detert and Burris (2007), building on Ashford et al., called an 'affect-…
- Distinguishing Voice and Silence at Work: Unique Relationships with Perceived Impact, Psychological Safety, and Burnout
Directly resolves a long-standing ambiguity that Kish-Gephart et al.
- Silenced by Fear: The Nature, Sources, and Consequences of Fear at Work
A theoretical synthesis arguing that existing accounts of workplace silence — built around a conscious, deliberate 'expectancy-like mental calculus' weighing the costs and benefits of speaking up — c…
- Employee Silence: Quiescence and Acquiescence as Responses to Perceived Injustice
The paper that split silence in two.
- Speaking Up to Higher-Ups: How Supervisors and Skip-Level Leaders Influence Employee Voice
An inductive, theory-building interview study (89 interviews across four units of a Fortune 500 high-tech firm, chosen as matched high/low pairs on a company 'speak up' survey) that reshapes how the…
- How We Respond
A corrective to the field's overemphasis on the speaking-up side of psychological safety.
- Safety Voice and Safety Listening During Aviation Accidents: Cockpit Voice Recordings Reveal That Speaking-Up to Power Is Not Enough
Speaking up is necessary and not sufficient, and this is the paper that demonstrates the second half at scale.
- Barriers to PS
Reports survey research (with Jade Garratt) on the experiential barriers to speaking up at work, based on 138 responses (62% including free-text commentary).
- On Face-Work: An Analysis of Ritual Elements in Social Interaction
The paper that turns face into an analytical concept for social science, and the source Edmondson (1999) reaches for when establishing that speaking up carries a cost to image.
- Language and Symbolic Power
A collection of essays arguing that language is not a neutral medium of communication but a site of power.
- Men and Women of the Corporation
Kanter's study of a large American corporation she called Indsco produced the theory of tokenism, and with it one of the most useful ideas in the whole literature on who can speak at work.
- The Vulnerable System: An Analysis of the Tenerife Air Disaster
Weick's analysis of the 1977 Tenerife runway collision, in which a KLM 747 began its take-off roll while a Pan Am 747 was still on the runway and 583 people died, is one of the founding texts of the…
- Least Safe Person
Argues a team is only as safe as the least safe person — and that measuring PS as an average actively hides the outlier who is suffering.