The Silence That Follows

Maria was twenty-six the first time, and she has not told this story at work in eleven years.

She had been at the plant three years and she had noticed something about the incoming inspection queue. It ran alphabetically by supplier name, which is how the system had always sorted it, which meant a supplier whose parts stopped the line on Monday got looked at after a supplier whose parts sat in a rack for three weeks. She wrote it up. Two pages, a quarter of queue data, one proposal: sort by line-stop risk instead of by name. No new system. A change to a sort order.

Her team leader thought it was good and sent it up. Somebody above him thought so too, because it went further, and for two weeks Maria heard things — it had been mentioned, someone had asked a question, it might go to the process council.

Then nothing.

Not a rejection. The precision matters here. Nobody said no, nobody said the analysis was wrong, nobody said it was somebody else’s decision to make. The thing stopped moving the way a stone stops moving, without any event you could point to.

She asked her team leader after a month. He said he would check. He may have checked.

At three months she asked again and he said something she still remembers: these things take a while. He was not being unkind. He was being accurate.

At six months she stopped asking. Nothing happened at six months either. There was no moment. She noticed one afternoon that she had stopped thinking about it, and that the noticing did not upset her much.

The queue is still alphabetical. Maria checked once, a few years ago, and has not checked since.

What she learned

The comfortable reading is that Maria was discouraged, that discouragement is a feeling, and that feelings pass. On that reading the fix is encouragement and the problem is morale.

That reading is wrong. Maria did not feel something. She learned something, and learning is not a mood. It is an update, and it outlasts the mood that accompanied it.

What she learned is that the relationship between offering an idea and anything happening was too weak to be worth the offering. Not that ideas get punished — hers wasn’t. Not that her manager was hostile — he wasn’t. Something more corrosive and much harder to argue with: that it does not matter.

And she was right. This is the part to sit with rather than skate past, because the temptation is to treat her conclusion as a distortion, a bit of learned pessimism to be corrected with better communication. It is not a distortion. She observed a system, formed a hypothesis about how it responds to input, ran the test once, and got a clean result. Eleven years later the queue is still alphabetical. Her model of the organisation is accurate. The organisation’s model of itself is the one that is wrong.

Now put her back in Chapter 1, thirty-seven years old, tool in hand, doing arithmetic about whether to raise it. I described that arithmetic as though she were performing it fresh — weighing IT, Legal, Quality Systems.

She wasn’t. She ran that calculation at twenty-six. Chapter 1 is not deliberation. It is recall.

The cost of staying down

Hiding is expensive, and it is expensive continuously. A book that only tallied the costs of speaking up would make the submarine look far more comfortable than it is.

Conservation of resources theory holds that people work from a finite pool of psychological resources, and that maintaining effortful states draws that pool down. Concealment is such a state. Maria keeps a private account, strips data before it moves, declines to expense eighteen euros, avoids explaining how she caught the seal defect early, and tracks, in every conversation about her work, which version of the truth she is currently in. None of it is difficult. All of it is constant.

The emerging work on covert AI use at work finds exactly this: hiding your AI use draws down psychological resources in ways that show up as strain and depletion. It converges with a much older finding about concealment generally, which is that keeping something hidden at work costs the hider something real, every day, for as long as it stays hidden.

So the submarine has a running cost. It is paid in attention and energy by the person crewing it and it never stops.

Which raises the question this chapter exists to answer. If hiding is expensive, and the organisation genuinely wants the idea, why does the expense not eventually push people up?

Because the alternative is worse, and worse in a way that lasts.

What the research calls this

There is a large, old body of work on this, and it sits nowhere near the innovation literature. That is one reason the innovation literature has been so bad at explaining the fleet. It sits in organisational behaviour, under employee voice and silence.

Elizabeth Morrison and Frances Milliken’s 2000 paper in the Academy of Management Review is the foundational statement. Their move was to argue that organisational silence is not an absence of something. It is a produced state — something organisations actively create through their structures and their managers’ beliefs, and then fail to notice they have created.

Three years later Milliken, Morrison and Hewlin interviewed employees about why they stayed quiet. The large majority had at some point felt unable to raise a concern. Only about half felt generally comfortable speaking up at all.

The useful part is not the prevalence. It is the breakdown of reasons. The most common was fear of being labelled — a complainer, a troublemaker, someone difficult. Close behind, damaging a relationship. And in third place, at roughly a quarter of cases, something different in kind: the belief that speaking up would change nothing.

Hold those apart, because the rest of the chapter turns on the distinction.

Fear is the belief that speaking up will hurt you. Futility is the belief that speaking up will do nothing.

From outside they look identical, since both produce a person who says nothing. They are not remotely the same thing.

Fear is the loud one

Fear is what everyone means when they discuss this problem, and the literature has treated it thoroughly.

Amy Edmondson’s work on psychological safety is the canonical treatment. Her 1999 study of hospital teams established the construct and showed that a team’s shared belief about whether it is safe to take interpersonal risks predicts whether people speak up, and downstream, whether the team learns. The construct has since accumulated meta-analytic support linking it to performance.

Chapter 2 gave us the sharpest version of why the fear is rational: Allen and colleagues found employees who used shadow IT were marked down whether their tool succeeded or failed. The penalty attached to the circumvention, not the outcome.

Fear is real, documented, and has a known remedy. The remedy is safety — leaders who respond well to bad news, teams where the interpersonal risk of speaking is genuinely low. Building it is hard, and Edmondson has spent a career on how, but the mechanism is understood and the industry has absorbed it.

Futility is the quiet one, and it does the damage

Here is the problem. Maria was not afraid.

Look for the fear in her story. Her team leader liked the idea and passed it up. Nobody labelled her, no relationship was damaged, and nothing in that account suggests she was taking a risk. By any reasonable audit she was in a psychologically safe environment. She went silent anyway.

Safety answers will this hurt me? Maria’s question was will this do anything? Safety does not answer the second question. It was never designed to.

Which is why well-intentioned interventions in this space fail while their sponsors remain genuinely puzzled. The most common is the open door — my door is always open, bring me anything. It is offered sincerely. It addresses fear. It does nothing whatever about futility, because the door was never the obstacle. What is on the other side of the door is the obstacle. An open door onto a room where nothing happens is not an improvement on a closed one. It is a more polite way of teaching the same lesson.

The same logic dismantles nearly every mechanism organisations reach for. Suggestion boxes. Innovation portals. Idea challenges. Skip-levels. Each is an invitation to speak, and each assumes people are silent because they have not been invited. Bernstein’s factory workers, from Chapter 1, were concealing better methods a few metres from suggestion boxes that stood empty. They had not failed to notice the boxes.

It compounds

If futility were only a mood, this would be an argument about tone. What makes it structural is that repeated exposure changes the person.

Andrieu and colleagues exposed participants to repeated instances of voice that produced no result — non-instrumental voice, in the literature’s phrase — and then measured not what those people said next, but how well they thought. Performance on a cognitive task fell. Motivation fell. Perceived control fell. What did not rise was any sense of being tired or distracted: the people affected did not report feeling worn down, they simply performed worse. Futility does not just close a mouth. It degrades the equipment.

That result sits on top of one of the most examined findings in twentieth-century psychology. Learned helplessness describes exactly this shape: repeated mismatch between action and outcome produces passivity, and the passivity generalises well beyond the situation that produced it. The organisational finding is new. The underlying mechanism is not.

Two further findings complete the picture.

Pinder and Harlos named the end state: acquiescent silence, distinguished from other kinds by the fact that nothing is being withheld. The person has given up hope of improvement and feels powerless to change things, so there is no suppressed idea to recover. Withheld silence still contains something. Acquiescent silence has stopped producing.

And Detert and Edmondson explain how the lesson persists without ever being re-taught. Their work on implicit voice theories found that people carry taken-for-granted beliefs about when speaking up is risky or pointless, and that these beliefs operate before any conversation begins. The person self-edits in advance. They do not weigh this idea against this manager, because the weighing already happened, generically, years ago.

Which is what Chapter 1 was. Maria did not evaluate her manager that Tuesday. She consulted a conclusion she reached at twenty-six.

The thing that should worry an executive

When a good idea is killed — or worse, allowed to stall — the organisation records the loss of one idea. Usually it records nothing at all.

What it lost is the person’s future ideas. Not this one. All of them. The person has updated, the update is about the class of action rather than the instance, and it does not expire.

Do the accounting slowly. Maria at twenty-six would have produced, over a career, some number of improvements of the kind she is capable of. She is a good engineer, she notices things, and she had eleven more years at that plant ahead of her. The alphabetical queue cost the company one sort order. The lesson cost it everything downstream of a person who concluded, correctly, that offering things is not how this works.

And none of it is visible. That is a structural property, not a rhetorical flourish, and it is why the mechanism has survived sixty years of people writing about innovation. There is no record of an idea that was never offered. No system captures it, no metric moves, nobody files a report saying I thought of something today and decided not to mention it. The rejections you can count are the visible tip. The extinction is invisible by construction.

Engagement surveys will not find it either, because the extinguished employee is not disengaged. Maria is excellent at her job. She catches seal defects early. She would score fine.

I think this is among the largest costs a large organisation carries, and it appears on no ledger anywhere. I am aware of the bind in asserting that: I have just spent three pages arguing the cost cannot be measured, which means I cannot produce a number for it either. I am asserting it anyway, because the alternative — declining to name a cost because the organisation has no instrument for it — is precisely the failure this book is about. An unmeasured cost is not a small one. It is an unmeasured one.

Why fixing the gate does not fix this

Here is the property that makes learned extinction the answer to why hasn’t anyone fixed it, and it is the hardest idea in the book.

Suppose you read Part II, accept every word, and go fix your gate. You separate two-way doors from one-way doors. You give increments a route that does not demand a business case. You defuse the territorial defence. You do all of it properly, and by Monday morning your organisation is genuinely better at receiving small ideas than it was on Friday.

Almost nothing happens.

The constraint was never only the gate. The constraint is a population that has already learned not to approach it, and that population does not re-learn on Monday because the gate changed. They will not test it. That is what the implicit-voice-theory finding means: the self-editing happens before the conversation, so the improved gate is never sampled. Your new process sits there, genuinely good, unvisited.

Engineers will recognise the shape. It is hysteresis — the state of a system depending not only on current conditions but on the path that got it there. Push the system into a new state and it does not return along the route it came. The damage persists after the cause is removed.

This is the best answer available to the question a sceptic should have been asking since Chapter 2. If all of this is so well documented, why has nobody fixed it? Partly because the mechanisms are individually reasonable, which was Part II. Mostly because the fix has a lag measured in years while the damage stays invisible, so nobody can tell whether their repair is working. An organisation that fixes its gate and sees no change within two quarters concludes the repair failed, reverts, and is wrong.

That is a hard thing to ask of an executive: do the expensive structural thing, receive no visible feedback for a long time, and trust that a mechanism you cannot measure is slowly reversing. I do not have a comfortable answer. I have a partial one in Chapter 12, and it is smaller than the problem.

Two mechanisms, one person

I have separated fear and futility cleanly, and Maria’s story isolates futility on purpose, because a mechanism is easier to see when nothing else is running.

Real cases are usually both. Chapter 1 Maria is plainly afraid — of IT, of Legal, of an eighteen-month programme that would find her tool embarrassing. Chapter 9 Maria, eleven years earlier, was not afraid of anything. Same person, two mechanisms, and the second one arrived after the first had already taught her the ground rules.

The separation is diagnostic, not natural. It matters because the remedies do not substitute for one another. Safety without responsiveness produces a comfortable room where nothing happens, which is the open door. Responsiveness without safety produces a manager who acts decisively on ideas nobody dares bring. Most organisations attempt the first and call it culture work.

There is also a population this chapter does not describe. Some people surface repeatedly, get nothing, and surface again. They exist in every company and you can probably name one. Seniority explains some of it, since a principal engineer can absorb a stalled idea that would extinguish a twenty-six-year-old. Temperament explains more. What I will say is that an organisation relying on those people is not solving the problem, it is being subsidised by the small minority whose response to futility is unusual. That is not a strategy, and it is not fair to them.

But staying down does not work either

I have spent a chapter arguing that surfacing is dangerous and that the damage from a bad surfacing lasts a career. Stop there and I would have written an argument for permanent submergence. So let me close that door, because it is the other half of the resolution.

The submarine fails too, on a longer timescale, for different reasons.

Maria’s tool exists in one place, which is Maria. It has no documentation, because documenting it means explaining it. It has no second user, because a second user is a witness. It runs on an account the company cannot see and does not know exists.

Nonaka’s work on organisational knowledge creation supplies the vocabulary. Knowledge that stays tacit — held in a person, expressed in how they work rather than in any artefact — cannot be transferred, combined or built on. An organisation’s capacity to create knowledge depends on tacit knowledge being made explicit and shared. Maria’s tool is tacit by design, and permanently so.

So when she transfers next spring, three things happen at once. The tool leaves. The seal defects go back to being caught five weeks later. And nobody connects the second to the first, because nobody knew about the first, so the degradation reads as ordinary variation.

This is the documented failure mode for improvement work generally. The continuous-improvement literature is consistent that local gains decay unless they are institutionalised — sustained through documentation, standardisation and organisational support rather than through one person’s commitment. There are widely quoted practitioner figures about how much decays and how fast. I am not repeating them, because I cannot trace them to anything solid and this book does not launder folklore into fact. The direction holds regardless: unsurfaced improvement is not permanent improvement. It is a loan against one person’s tenure.

Which gives the problem its real shape, and it is cruel.

Hiding costs the person continuously and loses the improvement eventually. Surfacing risks losing the improvement immediately and, done badly, costs the organisation every future improvement that person would have offered.

Both roads lose. The submarine is a bad permanent home and the surface is a dangerous place to be born.

Something has to give, and what gives is the assumption that this is one decision. It is two. The conditions an idea needs to survive its birth are not the conditions it needs to survive its adulthood, and the organisations that get this right stop asking people to choose between them.

That is Chapter 12. First we formalise the choice itself, and then we deal with the thing every executive wants to do the moment they finish this chapter, which is to go and find the fleet.

Where the evidence stops

The voice-and-silence research is decades deep and it is not in dispute. Morrison and Milliken established that silence is produced rather than absent. Pinder and Harlos identified the end state. Detert and Edmondson showed the self-editing runs before the conversation starts. Edmondson built the safety literature that explains the half of the problem everyone already knows about. Andrieu and colleagues took it a step further under controlled conditions, showing that repeated futility degrades executive functioning itself, sitting on a learned-helplessness foundation as solid as anything in psychology.

Every one of those describes silence. None of them describes hidden innovation. Twenty-five years of work on why people do not raise concerns, and almost none of it pointed at why people build things and say nothing — which is a different behaviour growing from the same root.

The hysteresis argument runs past the evidence, and it is worth being exact about where. The literature establishes that implicit voice theories form and that they operate pre-emptively. It stops there. Nobody has tested whether those beliefs update when conditions improve, which is precisely the question that decides whether a repaired gate ever gets sampled.

So here is the study I would like someone to run. Take an organisation that has measurably improved how it handles small ideas, and count the lag before submission rates move. If they move inside a quarter, the persistence argument in this chapter is wrong and Part III weakens with it. My expectation is that the lag runs to years.

Maria illustrates the mechanism. The research carries it. Those are different jobs and I have not confused them.

Next: Part IV. The dilemma, made rigorous — and why the curtain in Bernstein’s factory matters more than anything else in this book.