Silence in organisations is systematically misread. Leaders treat the absence of raised concerns as the absence of concerns — consensus, alignment, health — when the research record shows it is usually something else entirely: a rational calculation, run privately by employees, that speaking is not worth the price. Milliken and colleagues found that 85% of employees reported having felt unable, on at least one occasion, to raise an issue of genuine concern to their superiors — and that the reasons were neither timidity nor apathy but specific, learned expectations of futility and personal cost. This paper sets out silence as an economic phenomenon: a price mechanism operating on speech, set largely by leader behaviour in moments most leaders never notice. We examine what the silence of the informed costs organisations — drawing on public failures from Mid Staffordshire to the Columbia accident, where the decisive information existed inside the institution and did not travel upward — and set out the disciplines by which the price of speaking is deliberately lowered: not by declaring openness, but by changing what observably happens to the people who take the risk.
Begin with the finding that should end the conventional reading of quiet meetings. In their foundational study of employee silence, Milliken, Morrison and Hewlin found that 85% of the employees they studied reported at least one occasion on which they had felt unable to raise an issue or concern to those above them — issues spanning problems with processes, concerns about performance, and matters of fairness and conduct.1 These were not marginal employees, and the withheld items were not trivia; they were, in a large fraction of cases, precisely the operational and ethical information leadership most needed. The employees had views. They had run a calculation. The calculation said no.
This is the essential reframe: silence is not an empty state but an output — the product of a decision process that runs, mostly invisibly, in every employee before every potentially unwelcome utterance. Morrison and Milliken, mapping the phenomenon at the organisational level, showed that whole institutions can settle into what they called organizational silence — a shared, self-reinforcing understanding that speaking up about certain categories of things is not done here — sustained less by explicit prohibition than by structures and managerial signals that make the futility and cost of voice common knowledge.2
What does the private calculation weigh? The research converges on two terms, and their relative weight is itself instructive.
The first is risk: the expectation that speaking will cost the speaker — being labelled negative or a troublemaker, damaging relationships with the very people one must work with tomorrow, and, at the far end, formal retaliation. In Milliken and colleagues’ data, the fear of being viewed or labelled negatively, and the fear of damaging valued relationships, headed the list of reasons for withholding.1 Detert and Edmondson added a crucial refinement: much of this risk assessment is not a response to anything the current employer has done. People carry implicit voice theories — deep, early-learned rules such as “don’t embarrass the boss in public” and “don’t challenge upward without solid data and a solution” — which operate automatically, across employers, even under leaders who have done nothing to earn them.3 The practical consequence is sobering: a leader inherits silence they did not create, and neutrality will not dismantle it.
The second term is futility: the expectation that speaking will change nothing. In several strands of the voice literature, futility rivals or exceeds fear as the dominant suppressor — employees stay silent not primarily because they expect punishment but because they have learned, from watching what happened to previous input, that nothing happens.2,4 This matters because leaders working on psychological safety typically address the fear term (be warmer, don’t shoot messengers) while leaving the futility term untouched — input is received graciously and disappears — and then wonder why the safety scores rise while the silence persists.
The cost of silence is difficult to see precisely because its unit is the counterfactual: the error not caught, the risk not escalated, the idea not offered. Occasionally, however, an institution fails publicly enough that the counterfactual is reconstructed — and the reconstructions are remarkably consistent: the decisive information existed inside the organisation, held by people who had assessed, accurately, that voicing it was unsafe, futile, or both.
The Francis Inquiry into Mid Staffordshire NHS Foundation Trust — where failures of care were associated with substantial excess mortality over several years — found precisely this anatomy: staff who saw the failures did not speak, or spoke and were not heard, within a culture the report characterised as defensive and closed to concern; its recommendations centred not on clinical technique but on openness, transparency and candour — on re-pricing speech.5 The Columbia Accident Investigation Board, examining the 2003 shuttle loss, devoted much of its analysis to NASA’s organisational culture: engineers with concerns about the foam strike faced an environment in which, as the Board put it, the burden of proof had inverted — those worried about safety were required to prove danger, within reporting structures that muted minority technical opinion.6 Different sectors, identical mechanism: the information was present, and the price of moving it upward was set too high.
Edmondson’s original psychological-safety research supplies the everyday version of the same finding, with a twist that routinely misleads executives: in her hospital studies, the better teams appeared to make more errors — because they were the teams safe enough to report them.7 Silence does not show up in the data as silence. It shows up as clean dashboards, unanimous meetings, and surprises that were, in retrospect, no surprise to the people closest to the work. Google’s Project Aristotle, arriving from performance rather than safety, closed the loop: psychological safety — the shared belief that interpersonal risk is survivable — was the strongest single differentiator of its highest-performing teams.8
If silence is priced by evidence, where is the evidence generated? Overwhelmingly, in our observation, in a window most leaders never notice: the first seconds after someone says a hard thing. The flicker of irritation, the question answered before it is finished, the “let’s take that offline,” the subtle move to the next agenda item — each is, to the room, a price signal, broadcast to everyone watching how it goes for the person who spoke. Detert and Burris’s research on managerial behaviour and voice supports the mechanism: employees’ willingness to speak tracks their leaders’ observed openness far more closely than any stated policy.4 Nembhard and Edmondson identified the positive counterpart, leader inclusiveness — the active invitation and appreciation of others’ contributions — as the variable that unlocked voice across status gradients in clinical teams.9
Two implications follow. First, the price of speech is set behaviourally, not rhetorically — “my door is always open” is not evidence; what happened to the last person through the door is evidence. Second, because the assessments are updated on small samples, single events carry disproportionate weight in both directions: one visible punishment of a speaker re-prices speech for an entire division for years; equally, one visible episode of a leader thanking the bearer of unwelcome news, changing course on it, and crediting them publicly, is worth more than a year of values communication.
Work the futility term first. The fastest lever is not warmth but consequence: institute a visible loop in which raised concerns are logged, answered, and — where warranted — acted on, with the action attributed to the input. People speak into systems that demonstrably metabolise speech. Where input must be declined, decline it explicitly and with reasons; silence about input is itself input to the calculus.
Manage the three-second window as a practice. The discipline is trainable: when the hard thing lands, the leader’s first move is to receive it — thank, clarify, credit — and only then to evaluate. This is not softness; evaluation is not deferred, only sequenced. Teams we work with rehearse this explicitly, because the reflex under surprise is defence, and the reflex is exactly what the room prices.
Go and get the silence. Because implicit voice theories operate even under safe leaders, waiting for voice under-samples reality. The correction is proactive: directed questions to the people closest to the work (“what are we not hearing?” asked of named individuals with genuine room to answer), pre-mortems that license dissent by making it an assignment,10 and the deliberate soliciting of the most junior and most peripheral voices first in high-stakes reviews — inverting the status order that otherwise silences them.
Repair publicly when you get it wrong. Every leader will, at some point, punish a speaker — usually unintentionally, under pressure. The event is recoverable, but only in the same currency: a public naming of the miss (“I shut that down yesterday, and I was wrong to — here’s what I’ve done with the point”), which demonstrates that the price mechanism itself is being watched by the person with the most power over it.
A composite case: a national infrastructure operator, post-incident, scored well on engagement and openness surveys while its near-miss reporting ran at a fraction of sector benchmarks — clean dashboards, quiet meetings, and a frontline that, in confidential interviews, could name specific unreported risks and, in every case, the specific prior episode that had taught them not to report. The intervention was not a speak-up campaign (there had been three). It was the futility loop — every report answered within a fixed period, actions published monthly — plus rehearsed receiving behaviour for the top three leadership layers, plus one carefully handled public repair by the chief operating officer after an early lapse. Near-miss reporting tripled inside two quarters. Nothing about the workforce’s courage had changed. The price had.
If silence is an output of a price mechanism, it can be instrumented — and the instrumentation matters, because the standard instrument actively misleads. Engagement and culture surveys ask people whether they feel able to speak up, and people answer the question in general — against their most memorable experiences, under the same self-presentation pressures that produce silence in the first place. Hence the pattern in the infrastructure case above, which we see repeatedly: healthy speak-up scores coexisting with empty reporting channels. The survey measures the climate’s reputation. The behaviour measures its price.
Useful instrumentation therefore watches flows rather than feelings, and three flows carry most of the signal. First, upward problem flow: the volume and latency of concerns, near-misses, and bad news moving up each reporting line — not organisation-wide averages, which launder the variance, but per-leader, because the price of speech is set locally and a single expensive manager can silence a whole corridor behind healthy aggregates. Second, input consequence: what fraction of raised items receives a traceable response within the committed period — the futility term, made auditable. Third, surprise audit: for each significant adverse event, a structured look-back asking one question — who knew, and when? An organisation where post-incident reviews keep finding early internal knowledge has its answer about the price of speech, whatever the survey says.
Dimensional scanning of the kind described in White Paper 04 complements these flows at team level: asking not “is it safe here?” in general but where specifically the risk sits — dissenting from the leader, admitting error, delivering bad news upward, questioning how things are done — because the repairs differ, and the aggregate conceals which one is needed.
A brief note on conditions, parallel to the one in White Paper 04. Distributed work has changed the economics of voice in two opposing directions at once. It has lowered the cost of formal voice: written channels, asynchronous reporting, and the option to compose rather than improvise all reduce the exposure of speaking, and some organisations have seen reporting volumes rise on the move to hybrid. But it has quietly raised the cost of informal voice — the corridor version, the “have you got two minutes” version — which was always where the earliest, least-formed, most valuable concern first surfaced. A half-formed worry that would have been tested on a colleague at a desk now needs a meeting, an agenda slot, or a message that sits in writing — each a small escalation of commitment that the worry, at its earliest and most useful stage, does not yet justify. So it waits. The organisation experiences this as concerns arriving later and more fully formed — which sounds like maturity and is actually delay. The correction is deliberate: engineered informal surface area (standing unstructured time between levels, explicitly agenda-free) and leader behaviour that visibly rewards the early, unpolished raise over the late, complete one — because under distance, the early version no longer happens by accident.
Three limits. First, voice is not free even when safe: attention is finite, and an organisation that rewards all speech indiscriminately buys noise; the futility loop must include honest triage, transparently explained, or it collapses under its own volume. Second, the calculus is culturally inflected — the behaviours that signal safety, and the directness that voice may take, vary across cultures and status structures, and instruments must be calibrated accordingly; the underlying price mechanism, in our experience, does not vary. Third, safety is a property of climates, which are local: a safe team inside an unsafe institution protects its members imperfectly, and papers over none of the structural retaliation risk that formal whistleblowing frameworks exist to address. This paper concerns the everyday economics of speech; it is not a substitute for those protections.
For leaders, the model converts an abstraction into an inspection: What did the last three people to raise something difficult with you visibly experience? What happened, traceably, to their input? Those two answers are your speak-up culture; everything else is aspiration. For institutions — and for the public sector especially, where the Mid Staffordshire pattern recurs across inquiry after inquiry — the implication is that voice belongs in the risk register, priced and managed, not in the values statement. And for this desk’s larger argument: silence is one more case of the decisive organisational variable living in language — in what can be said here, and at what price — and therefore one more variable that leaders, once they can see the mechanism, can deliberately change.
IMPACT THINKING RESEARCH · BY BEN BOTES · WHITE PAPER 06 · 2026
The Psychological Safety Scan reads where it’s safe to speak and where it isn’t — in about three minutes.
Take the Psychological Safety Scan → Back to the research desk