Calling the Witness

Engels’ Pause, the Luddites, and a Proverb Nobody Studies

James Wolstencroft · July 2026

Testis Temporum · Phronesis · Paper I of III

Abstract

Every argument about artificial intelligence and the labour market eventually reaches for a historical comparison, usually the Industrial Revolution, usually in passing, usually to make a point about looms and then move swiftly on. This paper does not move swiftly on. It sits inside the one historical episode that already ran the experiment we are about to rerun: Britain between roughly 1770 and 1850, when a labour-replacing technology drove decades of GDP growth with no corresponding rise in real wages, a phase the economic historian Robert Allen named Engels’ Pause. It reframes the Luddite uprisings and the Captain Swing riots as a targeted political response to a specific institutional failure rather than as technophobia, and it identifies the two conditions, a shift in what the technology was for and a shift in who held bargaining power, that eventually ended the Pause. It closes by naming, precisely, the shape of the lesson this episode already tried to teach, so that the two papers following it can test whether that lesson has, this time, actually been learned. It has not, yet. That is what makes the proverb the series opens with worth the discomfort of repeating.

1. A pattern is not yet evidence

History is constantly teaching, but it doesn’t find many pupils.

That’s the proverb this series opened on, and the series has already put three independent witnesses on the record for it: a factory foreman from 1932, Santayana from 1905, Hegel from earlier still, none of them reading each other, all of them landing on the same complaint. Three independent corroborations is a pattern worth taking seriously. It is not, on its own, evidence, and I’d be committing the exact sin the proverb is describing if I admired the pattern, nodded at it, and moved on without checking whether it actually holds up against a real case. So this paper checks. It takes the one episode in economic history that already ran, in full, the experiment the rest of this series is about to rerun.

2. Engels’ Pause: growth without a wage

Between roughly 1770 and 1850, the British economy grew and the people making it grow did not get paid for the privilege.

The economic historian Robert Allen gave this phase a name, Engels’ Pause, after the young Friedrich Engels who watched it happen from inside Manchester’s cotton mills and wrote it up in 1845 as a condition of the working class rather than a statistic in a growth model. The data back him up uncomfortably well. Per capita GDP rose substantially across this period. Real wages for the people doing the work did not, for the better part of 80 years. That gap, output climbing while pay stands still, is a specific, diagnosable failure mode, and naming it matters. A problem without a name gets treated as weather rather than as a decision somebody made.

The decision, in this case, was about what the new machinery was for. Early industrial innovation (the spinning jenny, the water frame, eventually the power loom) was overwhelmingly labour-replacing rather than labour-enabling. It didn’t create new categories of skilled work that absorbed the people it displaced. It unbundled the tasks a skilled handloom weaver used to perform as one continuous, autonomous act of craft, and handed the unbundled pieces to centralised, low-skilled factory labour, operating under a discipline and a set of conditions the weaver’s workshop had never imposed. The surplus this generated (a genuine surplus, productivity really did surge) accrued almost entirely to the people who owned the machines. Profit’s share of national income rose. Labour’s share fell. The divide between what capital earned and what labour was paid widened every year the Pause continued, and it continued for the better part of a working lifetime.

3. The Luddites were not afraid of machines

Every account of the Luddite uprisings that reaches for the word “technophobia” has already gotten the story wrong.

Between 1811 and 1816, groups of English textile workers organised and, in a coordinated and specifically targeted way, destroyed industrial machinery. Popular memory has flattened this into superstition, a crowd of frightened artisans smashing looms because they didn’t understand them. The historical record does not support that reading, and it’s worth being blunt about why it persists anyway: it’s a much more comfortable story for everyone who came after. If the Luddites were simply ignorant, their defeat requires no reckoning. If they were making a specific, correct diagnosis of a specific, correctable institutional failure, their defeat requires an apology nobody has ever gotten round to issuing.

Because that is what they were doing. The machinery the Luddites targeted was destroyed selectively, not indiscriminately, aimed at the mills and the owners who were using it to bypass established wage protections and undercut trade customs that had, until that point, governed the relationship between a skilled worker and the price of his labour. This reads, at a glance, like a rejection of technology. It wasn’t, or rather it was something far more specific: a rejection of the institutional arrangement that let technology be deployed as a weapon against the bargaining position of the people it displaced, and an attempt to raise the cost of deploying it that way, direct and physical because no other channel was open to them. Fourteen years later, in 1830, the Captain Swing riots repeated the same targeted logic against the threshing machines then displacing agricultural labourers: coordinated letters, coordinated destruction, coordinated demands. A movement, not a mob, with a specific institutional grievance and no legal mechanism through which to press it.

Governments had a choice at this point, and they made it quickly. They did not respond by addressing the underlying inequality, by asking why productivity gains were flowing entirely to capital while the people producing them starved. They responded with repression: the criminalisation of labour organising, and in England, by 1812, a statute that made machine-breaking a capital offence, punishable by death. The state deployed military force, more troops in some years than were fighting Napoleon on the Peninsula, to protect capital assets from the people capital had just made obsolete. That is the whole shape of the failure, stated plainly, and it’s no small detail: the correct diagnosis was delivered, by the people living inside the problem, and the response was to hang the diagnosticians.

4. What actually ended the Pause

Real wages only decoupled from stagnation after 1850, and they decoupled because two separate things changed at once, not one.

This is the part of the story that gets dropped when the Industrial Revolution is invoked as a reassurance: the “it worked out in the end, didn’t it” argument, reached for by people who want permission to stop worrying about AI. It did work out. Eventually. After roughly 80 years, which is longer than most of the people currently reassuring you will live. And it worked out for two specific, separable reasons, and it is worth insisting on the separateness, because conflating them is exactly how the reassurance gets misapplied to the present.

The first reason is that the technology itself changed character. Innovation after 1850 increasingly created entirely new sectors, new occupations, and tasks that complemented human labour rather than simply replacing it (engineering, management, an expanding range of skilled trades the earlier machinery had not required and could not itself perform). The second reason is institutional, and it is the one that keeps getting quietly dropped from the “it worked out” version of the story. Legal labour protections expanded. Trade unions, illegal or barely tolerated for much of the Pause, grew into a real collective bargaining force. Political enfranchisement widened, giving the people bearing the cost of industrialisation an actual electoral lever over the people setting its terms. Neither of these two changes, on its own, would have closed the gap. Labour-enabling technology without institutional bargaining power still lets capital capture the gains; ask anyone in a sector where the union was broken before the new machinery arrived. Institutional bargaining power without labour-enabling technology has nothing to bargain over except a shrinking pie. It took both.

5. Naming the lesson, precisely

The lesson Engels’ Pause tried to teach sounds like it should be “technology causes suffering.” It isn’t, not quite; it’s more specific than that, and specificity is the whole point of writing it down.

The lesson is this: market forces alone do not guarantee that the gains from a labour-replacing technology reach the people that technology displaces. Left to itself, a labour-replacing shock will route its surplus toward whoever already owns the capital, for as long as the institutions surrounding it allow that to happen, and the institutions surrounding it will allow that to happen for exactly as long as the people bearing the cost lack the organised power to force a different settlement. That claim is mechanical rather than moral: it describes where surplus flows in the absence of a countervailing force, whatever the intentions of mill owners or AI labs. Mechanical claims are testable, which is precisely why this paper has spent five sections building the test case rather than the sermon.

I want to be honest about what this test case cannot do on its own, because a historical parallel that oversells its own certainty is worse than no parallel at all. Engels’ Pause is one episode, in one country, in one century, and every economic historian who has studied it, Allen included, has argued at length about exactly how much of the wage stagnation to attribute to technology as opposed to war, population growth, or the Corn Laws. Analogy is not proof. What a well chosen historical episode buys you is a set of questions worth asking, sharpened by having already watched them play out once, not certainty about the present. Is the current wave of labour-replacing technology creating labour-enabling work fast enough to offset what it displaces? Do the people being displaced currently have anything resembling the institutional bargaining power that took 80 years and a great deal of blood to build the first time? Those questions are the entire subject of the paper that follows this one.

Limitations and open problems

This paper has deliberately stayed inside one country and one century, and it owes the reader an account of what that costs. First, it treats Engels’ Pause as a single, coherent episode with a legible cause and a legible resolution, when serious economic historians still argue about how much weight to give technology against war finance, population pressure, and the Corn Laws in explaining the wage stagnation; the reading given here is the mainstream one, not an uncontested one. Second, it uses the Luddites and Captain Swing as evidence of institutional failure without claiming they were tactically successful; they lost, badly, and a lesson delivered by the losing side is not automatically the correct one merely because it turned out, decades later, to have been vindicated. Third, and most importantly for what comes next, this paper does not attempt to establish that generative AI behaves like a nineteenth century power loom. It establishes what to look for. Whether the pattern actually holds today, in the numbers rather than the analogy, is the entire task of the next paper in this series, Lux Veritatis, and it does not get to inherit this paper’s conclusions without earning them again from scratch.

Boundary notes

Testis Temporum supplies the precedent only: what happened, why it happened, and what ended it. It makes no claim about generative AI’s present labour market effects, which belong entirely to Lux Veritatis, and no claim about what policy should follow, which belongs entirely to Magistra Vitae. Readers tempted to treat Engels’ Pause as a settled prophecy for the 2020s should read Section 5’s caveats before they read the numbers in Lux Veritatis.

References

• Allen, R. C. “Engels’ Pause: Technical Change, Capital Accumulation, and Inequality in the British Industrial Revolution.” Explorations in Economic History, 2009.
• Allen, R. C. “Technical change, globalisation and the labour market: British and American experience since 1620.” Institute for Fiscal Studies / Deaton Review.
• Frey, C. B. The Technology Trap: Capital, Labor, and Power in the Age of Automation. Princeton University Press, 2019.
• Clark, G. and Cummins, N. “Inequality and social mobility in the Era of the Industrial Revolution.”
• Engels, F. The Condition of the Working Class in England. 1845.
• Hegel, G. W. F. Lectures on the Philosophy of History. Translated by J. Sibree.
• Santayana, G. The Life of Reason. 1905.
• Myers, H. and Roberts, M. M. Human Engineering. 1932.
• Quote Investigator. “You Must Learn from the Mistakes of Others.” 18 September 2018; “Those Who Cannot Remember the Past Are Condemned To Repeat It.” 4 March 2024; “We Learn From History That We Do Not Learn From History.” 6 March 2024.
• Cicero, M. T. De Oratore, Book II.