Nobody Learned to Speak From a Book.

Why behaviour change programmes don't stick — and what human beings have always known about how behaviour actually spreads.

You have an accent. Nobody taught it to you.

No one sat you down at four years old and explained the vowel shifts of your region. There was no module. There was no handout. You acquired one of the most intricate motor skills a human performs — thousands of muscle movements per minute, coordinated to a tolerance of milliseconds — entirely by being in a room with people who were already doing it.

The same is true of almost everything you do. How close you stand to someone. What you do with your hands when you're nervous. Whether you interrupt. How you hold silence. When you look away. Nobody taught you any of it, and you could not write down the rules if you tried. You copied.

This is not a charming fact about childhood. It is the primary mechanism by which our species transmits behaviour — older than writing, older than schooling, and still running at full power in every corridor of every organisation on earth. Which raises an uncomfortable question for anyone whose job is changing how people behave at work:

If human beings have never learned behaviour from instruction, why do we keep trying to change it with courses?

What the children did to the doll.

In 1961, Albert Bandura put a group of nursery-age children in a room with an adult and an inflatable clown.

Some children watched the adult play quietly. Others watched the adult attack the doll — striking it, kicking it, shouting at it. Nobody told the children to do anything. Nobody praised the adult or rewarded them. The children simply saw.

Then each child was left alone with the same doll. The ones who had watched aggression produced it — not roughly, but precisely. Same strikes. Same phrases. Some of them invented new attacks in the same style, having absorbed not just the actions but the logic of the actions.

Bandura's conclusion founded social learning theory and it is now so absorbed into common sense that it sounds banal: people acquire behaviour by observing other people, without instruction, without reward, and often without any awareness that learning has occurred.

The study has been argued over for sixty years — about the novelty of the doll, about what it does or doesn't prove regarding media violence. The core finding has survived all of it, and has been replicated in forms Bandura never used.

And every parent knows it in their bones anyway. It's why we lower our voices when the children are in the room, and why we feel caught out when a four-year-old repeats something we said in traffic. We understand instinctively that we are not teaching our children with our words. We are teaching them with ourselves.

The strangest thing about human copying.

Here is where it gets genuinely odd, and where the implications for organisations become sharp.

In 2005, Victoria Horner and Andrew Whiten ran a study comparing young chimpanzees with three- and four-year-old children. Both watched an adult retrieve a reward from a puzzle box. The demonstration deliberately included a number of steps that were completely pointless — tapping the box, poking at a section that did nothing — before the actions that actually worked.

When the box was opaque, chimps and children both copied the whole sequence. Fair enough; they couldn't tell which bits mattered.

Then the researchers made the box transparent, so the useless steps were visibly useless.

The chimpanzees dropped them immediately and went straight for the reward. Efficient, rational, sensible.

The children kept doing them. All of them. Every pointless tap, faithfully reproduced, even though they could plainly see it achieved nothing.

This is called over-imitation, it's been replicated across cultures, and — this is the part that should stop any L&D professional in their tracks — it has been found in adults too, at even higher fidelity than in children.

We are not efficient copiers. We are faithful copiers. When we watch someone who matters do something, we absorb the whole ritual: the useful parts, the pointless parts, the defensive parts, the bits that made sense in 1997. We copy the structure of what we see, not the parts that work.

Which means an organisation doesn't just inherit its leaders' competence. It inherits their flinches.

The channel that's always on.

Now hold those two findings together and look at how organisations try to change behaviour.

Somebody notices that managers aren't giving honest feedback. So a programme is commissioned. There's a model with four quadrants. There are role-plays in a hotel outside town. There are slides, a workbook, a follow-up webinar, and a satisfaction score of 4.6 out of 5.

Then everyone goes back to a building where the executive team has not given each other honest feedback in eleven years.

Which curriculum do you think wins?

The programme runs for two days. The demonstration runs for two hundred and forty days a year, at close range, performed by the people who decide promotions. It is delivered in the native format of human behavioural learning — watching someone who matters do a thing and survive — while the programme is delivered in the one format our species has never used for this purpose: description.

Chris Argyris gave us the language for the gap fifty years ago. Espoused theory is what an organisation says it values: candour, challenge, psychological safety. Theory-in-use is the set of rules that actually govern behaviour, visible only in what people do. Argyris found that almost nobody can accurately state their own theory-in-use — and that organisations systematically reward the defensive routines they publicly deplore: saving face, avoiding embarrassment, keeping the difficult thing unsaid.

Training teaches the espoused theory. The corridor demonstrates the theory-in-use. And people copy what is demonstrated, because that's the only channel their behavioural learning system was ever built to receive.

We're not failing to teach the behaviour. We're teaching two contradictory behaviours through two different channels, and being surprised when the stronger channel wins.

What the research actually says about decay.

L&D professionals have long had a statistic for this: only 10% of training transfers to the job. It appears in board papers, keynotes and business cases everywhere.

It's folklore. Robert Fitzpatrick traced it in 2001 and found it originated in an off-hand estimate in a 1982 article by Georgenson — no study, no data, no methodology. It survived four decades of citation because it felt true to everyone who read it. Which is, in itself, a finding worth sitting with.

The properly researched picture is less quotable and more serious. Reviews of the skill-decay literature suggest that somewhere between half and nine-tenths of what people acquire in training has degraded within a year. And the loss begins immediately — Hermann Ebbinghaus established the shape of it in the 1880s, testing his own memory: the steepest fall happens in the first hours and days, not the last months. Baldwin and Ford named the field's central puzzle in 1988 and called it the transfer problem. Nearly forty years on, it is still the problem.

But notice that "transfer" already contains the wrong assumption. It imagines behaviour as a substance that moves from a classroom into a workplace, and asks why so much is lost in transit.

Nothing is being lost in transit. The behaviour is being overwritten on arrival — by a room full of people demonstrating something else.

Why the defensive patterns are the stickiest of all.

There's a further turn, and it's the one that matters most.

The behaviours organisations most want to change are almost never neutral skills. They're defensive: smoothing over the disagreement, deferring upward, going quiet, shutting the conversation down, defending the position, taking over, firefighting.

These are the hardest to shift for three compounding reasons.

They're fast. Joseph LeDoux's work on threat processing found a direct route to the amygdala that fires in roughly twelve milliseconds — while conscious awareness of the same event takes twenty times longer. Under pressure, the body has committed before the framework from module three has even been retrieved. You cannot out-teach a reflex that arrives first.

They're demonstrated by the powerful. Defensive behaviour is most visible in senior people under pressure — which is precisely the condition in which juniors are watching most closely, because that's when they're learning what survival looks like here. The over-imitation finding says they won't filter it. They'll copy the whole ritual.

They work. Smoothing does dissolve the tension in the room. Taking over does get the deliverable out. Every defensive behaviour carries a genuine gift — care, standards, harmony, decisiveness, urgency — and delivers a real short-term win. That's exactly why they propagate. Nobody copies a behaviour they watch fail.

So the pattern isn't decaying through neglect. It is being actively re-taught, every day, by the most credible teachers in the building, in the format humans actually learn from, and confirmed by results.

Against that, a workbook doesn't stand a chance.

What this means for L&D.

None of this is an argument against training. Skills need teaching and teaching works — for skills. Instruction is a superb channel for knowledge, tools, systems, technical craft.

It's an argument against a category error: treating behaviour as though it were knowledge. You cannot instruct your way to a behaviour that nobody in the building is modelling. All you achieve is a workforce that can describe, fluently and sincerely, the thing it still doesn't do.

Three things follow.

One: the real curriculum is the leadership team. Not what they say — what they do when the numbers are bad. That's the syllabus everyone is actually studying, and no programme can outvote it. This is why behaviour change is not a volume problem. It's a leverage problem. Change what a small number of watched people do under pressure, and you have changed what an entire organisation is learning, without teaching anyone anything.

Two: those people don't need instruction, they need release. A senior leader's defensive pattern isn't an information gap — they can usually describe the better behaviour better than you can. It's an old protection running on automatic. It loosens through sight, not through study: when someone sees the pattern whole — what it was built for, what it earned them, which threat it's still guarding against — the automatic becomes optional. That's a different kind of intervention from a course, and a much faster one, because insight isn't a curriculum and doesn't need twelve months of scaffolding.

Three: sequence beats volume. Release first, then train. Run the same feedback programme after the modelled behaviour has changed, and suddenly the content has somewhere to land — because now it describes what people can already see being done.

And measure the right thing. Not attendance, not satisfaction sheets, not self-reported confidence. The honest question is whether the conversations that weren't happening are happening now. Crucial Learning prices a single avoided conversation at around $1,500 in workarounds and wasted effort — and no organisation avoids just one.

Back to the accent.

You learned to speak by being in the room with speakers. You learned to argue by watching people argue. You learned what to do when someone senior is angry by watching what the adults around you did when someone senior was angry — and you learned it so thoroughly that you can't remember learning it at all.

Nothing about that mechanism switched off when you got a job. It is running right now, in every meeting in your organisation, transmitting the real behaviour at full strength while the programme transmits the official one at a whisper.

We've always known this. It's why we watch our language around children, why apprenticeships outlast textbooks, why every culture on earth taught its young through story and demonstration rather than instruction. It isn't new. It's the oldest thing we know about ourselves.

We just stopped believing it the moment the learner started wearing a lanyard.