Sunday, November 06, 2011

Steven Pinker and the An Lushan Revolt

Most people – I think – if asked to name the bloodiest century in human history would probably say the 20th. I hasten to add this isn't the kind of question you get on history exam papers nowadays – you are more likely to get quizzed on ‘Household Formation, Lineage and gender relations in the early modern Atlantic world’ or something less bloodthirsty.

In Steven Pinker’s new book, ‘The better angels of our nature’ the wild haired Harvard professor is having none of this. The preference for the 20th century is mere ‘historical myopia’. Instead when one roots around through the
history books for forgotten wars and scales for the world’s population at the time – you find a whole set of lesser known conflicts that dwarf the toll for the first and second world wars. Pinker then presents a table showing the Second World War as merely the 9th most destructive atrocity of all time – lagging behind the Atlantic Slave Trade, the annihilation of the American Indians, Tamerlane’s conquests, the fall of Rome, the fall of the Ming dynasty, the Mid-east slave trade, the Mongol conquests and – most terrible of all – the An Lushan Revolt (something the majority of westerners have never even heard of).

Now at this point one’s proverbial ‘Bullshit-o-meter’ should be sounding – anyone who claims that they have a reasonably accurate ‘death toll estimate’ for something like the Mongol Conquests is being ludicrously over-confident. Pinker’s table looks suspiciously like something that has been cut and pasted from Wikipedia. In fact the figures appear to have been lifted from a site called ‘Necromterics’ authored by Matthew White – a librarian and author whose somewhat macabre hobby appears to be calculating historic death tolls. His scholarly works include such essays as 'Which Has Killed More People? Christianity? or Gun Control' so it's a bit strange that Pinker would consider him the go-to man on the demography of Medieval China.

The An Lushan Revolt, according to Pinker and White, wiped out something like 36,000,000 Chinese over the course of 8 years – a toll equivalent to two thirds (66%) of the Tang Empire’s population. If you scale for the mid 20th century’s population you would end up with an equivalent toll of 429,000,000 people. That would indeed be an astonishing high death rate – by comparison the Nazi invasion of Soviet Russia killed around 13% of Russia’s population - over half the population in the regions and countries of Europe where there is data of useful quality died in the Black Death (perhaps the worst demographic disaster in the history of the world). To justify this Pinker and White refer to the fact that at the peak of the medieval Tang dynasty, the census taken in the year 753 recorded a population of 52,880,488. After eleven years of civil war, the census of 764 gave a figure of 16,900,000. None of the figures cited on White’s site appear to come from Sinologists as far as I can see and no context is given for the low census figures*.

Accordingly I have worked through a number of works such as the ‘Cambridge History of China Vol 3’, Mark Edward Lewis’s ‘The Chinese Cosmopolitan Empire – the Tang Dynasty’ and David Andrew Graff’s ‘Medieval Chinese Warfare’ to see if they can shed greater light on what is now claimed to be the greatest holocaust in human history.

An Lushan was a garrison commanding general of mixed Sogdian and Turkish descent who rebelled against the Tang Dynasty in 755. This sparked a civil war across northern China for a period of eight years before the rebels were finally destroyed in 763. During this period of the two capitals of the Tang dynasty, the city of Chang’an was damaged and the city of Luoyang was burned. This suggests that the conflict was highly destructive but when assessing the impact there are a number of difficulties.

Firstly, up until the modern age, population counts were sporadic and incomplete. The first full censuses were not made until 1790 in the United States and 1801 in Britain. In the medieval Chinese era, the government counted households and some or all of the people constituting them, but did not attempt a complete registration until 1953. This was for the purpose of levying troops or more commonly allocating tax burdens. Only a few landmark censuses from the pre-Song era are taken to be reasonably reliable and the taxation records are frequently disrupted by war and administrative chaos. The figures for number of households are held to be far more reliable than those for actual head count

Secondly the census figures vary wildly depending on the contemporary level of government control. For example, in the reign of Taizong from 626 to 649, only 3,000,000 household were registered. Under the previous Sui dynasty (581-618) the figure had been 9,000,000 households. According to Richard Guisso in the Cambridge History of China this ‘sensational decline was not the result of catastrophic loss of life during the civil warfare of late Sui and early T'ang, but of simple failure by the local authorities to register the population in full. Even in the first years of Kao-tsung's reign only 3,800,000 households - certainly far less than half of the actual population – were registered. Considerably more than half of the population was thus unregistered and paying no taxes (p297 Cambridge History of China). This shows that in times of difficulty the highly centralised taxation system could break down – resulting in half the population or more being omitted from the census.

After the An Lushan revolt the situation reached crisis proportions and a new period of warlordism and regional autonomy emerged. The Tang had survived only by carrying out a general decentralisation of administrative power and dispersing power through a new tier of provincial governments. Despite the restoration of peace the empire remained in a state of chaos. China broke into many regions who collected their own taxes and remitted only a small portion to the central government. The Tang could no longer update it’s registers and chart landowning; local tax records were destroyed, scattered and rendered obselete. As Graff writes:

After the An Lushan rebellion, the Tang court lost the ability to enroll, enumerate, and impose taxes directly upon the majority of China’s peasant households. This development is dramatically illustrated by the decline of the registered population from approximately nine million households in 755 to less than two million in 760. (P240 – Medieval Chinese Warfare)

The post rebellion census figures cannot then be relied upon when estimating the impact on the empires population in the 8th century and there are no signs of a catastrophic two thirds population loss. Instead the indications are that China continued to have a large population base into the 9th century with which the dynasty was able to raise professional and conscripted armies to compete with the nomadic powers in dominating inner Asia.

The estimates given by the great Harvard sinologist John King Fairbank in 'the New History of China' (2006) are that ‘the empire’s population may have totalled 60 million in AD 80, 80 million in 875, 110 million in 1190’ (p106). These are of course estimates but they show that the general impression from historians of the period is not one of catastrophic population decline followed by recovery – but of a slow and steady late medieval population boom coupled with a shift in population from north to south. Mark Edward Lewis remarks that that:

‘Between 742 and 1080 (two years for which comprehensive census records have survived), the population in the north increased by only 26 percent, while that in the south increased by 328 percent’

C A Peterson in the Cambridge history of China notes that in the wake of the rebellion:

Large scale shifts of population took place. Many of the war affected areas in Ho-pei and Ho-nan were partially depopulated, and many people migrated to the Huai and Yangtze valleys and to the south (P496)

There are therefore plenty of reasons to be sceptical of Pinker’s claim that An Lushan’s revolt ranks as the most destructive war of all time. In fact he doesn’t appear to have done even the most basic research of research into the credibility of his figures; which is a shame because ‘The Better Angels of our Natures’ is a very good read and presents some interesting questions.

*In his recent book 'The great big book of horrible things' Matthew White goes with a 'more conservative' figure for the An Lushan revolt of 13 million dead - though he obviously didn't tell Pinker.

Discuss this post at the Quodlibeta Forum


Anonymous said...

Thanks very much for fact checking Pinker's claim. I fear this won't be the only historical howler in his book. When scientists dabble in history, the results are often quite brutal and Pinker seems to be no exception.

Humphrey said...

I've got enough howlers for a whole series - in fact this blog might have to be changed from 'Quodlibeta' to 'The Steve Pinker Trolling Blog'.

Doug said...

please keep them coming!

Ignorance said...

"In his recent book 'The great big book of horrible things' Matthew White goes with a 'more conservative' figure for the An Lushan revolt of 13 million dead - though he obviously didn't tell Pinker."

That's weird, because there is an edition of that book with a foreword by Pinker. You'd think he would have read that figure.

Keen Reader said...

If only these New Atheists would learn to think for themselves . . .

Anonymous said...

Keep it up.

Anonymous said...

What has any of that got to do with understanding the state of the world altogether in 2011?

Please find a completely different very sobering assessment of the now-time world situation altogether altogether via these two references.

Anonymous said...

None of the figures cited on White’s site appear to come from Sinologists as far as I can see and no context is given for the low census figures*.

The first half of the sentence seems to me unfair to White. The second half explains the real problem (detailed in the rest of your post), but the first half implies that his numbers are wrong. Maybe they are, but the quote from Graff is quite compatible with White's numbers.

Also, "suspiciously like something that has been cut and pasted from Wikipedia" is unfair to Wikipedia, which explains the context!

Gyan said...

Pinker is simply rabidly anti-Christian
and is willing to fudge, hedge and misrepresent facts to any extent possible.

He should not be considered a serious and respectable scholar at all.

Just see his FAQ where he writes that Vatican did nothing against Hitler.

Humphrey said...

'The first half of the sentence seems to me unfair to White'

I should point out White has hedged his bets somewhat. On his site he has An Lushan down as causing 10 million deaths. In his book he has it down as 13 million. The number Pinker seems to have got off him is 36,000,000 (?!)

On the sources point - yes, it's a bit unfair. The sources for his figures do appear to be encyclopedias, however there are some references from journals - e.g population studies which do refer to the work of Sinologists. However in the case of An Lushan these say:

'Many historians have affirmed that 36 million lives were lost as a result of the violent event, but Fitzgerald and others have shown that this is incredible. Even if such a huge loss were conceivable, it would be naive to suppose that an accurate count could be carried out in the midst of the ensuing chaos. Fitzgerald (Charles Patrick Fitzgerald), 1973, p.312-315 ("The real cause of the decline in the figures for the censuses after the rebellion was the dispersion of the officials who had been in charge of the revenue department.")'

That would argue for the skeptical end of the spectrum - not the figure Pinker was given.

Anonymous said...

YOu have cited an impressive array of AUTHORITYS to counter pinker, but have not cited what evidence these gentlman have - after all, these scholars are living in Western Europe and the US hundreds of years after the fact; from your long list of quotes, it is not clear how much weight should be given to these assertions - it is possible (logically) from what you have written that every single author you cite is relying on teh same source...

Humphrey said...

Well - they are all relying on the same source; namely the census records. The issue is how the source should be interpreted, does it show that 2 thirds of the empire was wiped out or does it show a loss of government control ? I don't have the expertise to answer that question but Sinologists who have spent their lives looking at the period do - and yes considerable weight should be given to their assertions, above and beyond that of amateur hobbyists.

Mark Frank said...

Just so we are clear - we are talking about one line in a table in a book which is stuffed with statistics and facts and Pinker says of the table:

"These figures cannot of course be taken at face value. Some tendentiously blame the entire death toll of a famine or epidemic on a particular war, rebellion or tyrant. And some come from innumerate cultures that lacked modern techniques for counting and record keeping."

Anonymous said...

which makes the attempt to base an argument on these numbers all the more asinine and intellectually dishonest.

Mark Frank said...


Have you read the book?

Anonymous said...

Some of it (life is short). What is your charge?

Mark Frank said...

I just wondered what argument you thought he was basing on these numbers?

Anonymous said...

The argument that the 20th century was considerably less violent and destructive than, say, the Middle Ages.

Humphrey said...

Hi Mark - he does add that caveat but then carries on in the same vein:

"...At the same time, narrative history confirms that earlier civilizations were certainly capable of killing in vast numbers. Technological backwardness was no impediment; we know from Rwanda and Cambodia that massive numbers of people can be murdered with low-tech means like machetes and starvation. And in the distant past, implements of killing were not always so low-tech, because military weaponry usually boasted the most advanced technology of the age. The military historian John Keegan notes that by the middle of the 2nd millennium BCE, the chariot allowed nomadic armies to rain death on the civilizations they invaded. “Circling at a distance of 100 or 200 yards from the herds of unarmored foot soldiers, a chariot crew—one to drive, one to shoot—might have transfixed six men a minute. Ten minutes’ work by ten chariots would cause 500 casualties or more, a Battle of the Somme–like toll among the small armies of the period. High-throughput massacre was also perfected by mounted hordes from the steppes, such as the Scythians, Huns, Mongols, Turks, Magyars, Tatars, Mughals, and Manchus. For two thousand years these warriors deployed meticulously crafted composite bows (made from a glued laminate of wood, tendon, and horn) to run up immense body counts in their sackings and raids. These tribes were responsible for numbers 3, 5, 11, and 15 on the top-twenty-one list, and they take four of the top six slots in the population-adjusted ranking. The Mongol invasions of Islamic lands in the 13th century resulted in the massacre of 1.3 million people in the city of Merv alone, and another 800,000 residents of Baghdad"

So he is clearly adhering to the viewpoint that - while the figures may have considerable margin for error one can nevertheless use them to conclude that events such as An Lushan and the Mongol Conquests were comparable to World War II (when adjusting for world population). Now that is an interesting argument and one which deserves further analysis which is what I have tried to do.

Now a few have said that this is only one table in a book full of arguments; which is undoubtedly true. However I found it was one which reviewers really focused on (E.G - ). It was also something Pinker majored on in the pre-release publicity and was one of the things that made me want to buy the book in the first place - hence why I have focused on it. I don't think it is peripheral to his overall thesis as you and others have made out since it seems to be a key pillar of his view that a decline in violence has taken place.

Humphrey said...

Pinker summerises his arguments here:

The graph and White's statistics come in at about 6.00 - they are crucial to the idea of 'the long peace'.

Anonymous said...

I second Humphrey's reply and would have written something similar (albeit of much lower quality) if I could have been bothered. The bottom line is that Pinker, even if he is aware of the lack reliability of his statistics, desperately needs them to create pre-modern contenders for the 20th century World Wars. And this is where his approach is fundamentally dishonest.

Mark Frank said...


First – I should say that while I think the book is a rather brilliant polemic I do accept that some of the evidence is questionable. I don’t think it is an academic text. It is a popular book designed to present a different way of thinking about our current state. I should add that while the book indicates where religion has been responsible for violence its main thrust is not anti-religious. I could imagine a very similar book written by a believer.

Some specifics:

* Let's stick to the book. It is very easy to miss out or misrepresent things when making a speech.

* The whole thing about whether the 20th century is the worst for violence is a just a prelude to one chapter of the book – something he wants to get out of the way before going on to the meat of the chapter. He is arguing that there have been other periods of comparable violence and making the point about historical myopia. Do you disagree with either of these conclusions?

* The White table is only part of the evidence for other centuries being as bad as the 20th. As you yourself point out he supplements this with qualitative evidence. The An Lushan rebellion is just one item in this table and is never referred to again in the whole book as far as I can see.

* Pinker makes it quite clear that the White data in general and the An Lushan revolt in particular are open to dispute (if you had looked at his footnote about the An Lushan revolt you would hardly need to have done your research).

* While most of the rates in the White table seem to be too high, and sometimes implausibly high, Pinker didn’t need to exaggerate (and I wish he hadn’t). The quoted An Lushan rate is eight times the rate for the second world war. His point is made even if it is a fraction of what he wrote.

Anonymous said...

I'm really terribly sorry, but all I seem to be getting out of your defence is that "throwing numbers against the wall till something sticks" is an acceptable method in science and scholarship. Is that really what you're saying? According to my reading of your defence of Pinker, I could pull any ridiculously inflated number out of my behind in support of any given argument and would always be right, because my case would be supported even if it happened to be "just a fraction". Yet the point Humphrey was trying to make with his post is that there is no good reason, based on the available evidence, to think that it was even that relevant fraction.

Granted, the An Lushan revolt plays a minor, perhaps negligible, part in his book, but nobody disputes that. And you must concede that it is one of the "factoids" that are being eaten up by his audience, precisely because of its sensationalistic nature.

I would like to emphasize that I don't care whether or not Pinker tried to badmouth religion - an intent that I consider perfectly legit. What I care about is the soundness of his or anybody else's scholarship. Why would we let Pinker get away with stuff that would lead to negative repercussions for any PhD candidate. To call something a "popular book" is more an indictment than an exoneration in my view.

PS: You are talking about earlier periods being "as bad as" the 20th century. But that would not be enough for Pinker's general argument.

Humphrey said...

Hi Mark – agree with the first bit. Anything that presents a new way of looking at the past and correcting misconceptions is welcome; that said I think it could have been done a lot better. As a compilation of statistical arguments it works well (for example it led me to Lawrence Keeley work on hunter gatherers in ‘War before Civilization)

On the anti-religious comment – nothing wrong with being anti-religious but some of the statements Pinker makes are just flat out wrong e.g saying the Crusader massacres were equal in relative quantity to the holocaust. No-one these days would want to defend the Crusades (well, except for Rodney Stark and some US conservatives) but one has to keep things in perspective.

Stick to the book – yes true but the fact he keeps putting the emphasis on that table and referring to the data from it in the rest of his discussion does mean we should give it some weight. Then there is also the impression people get from the book – the ‘take home message’. I see a lot of people in reviews referring to An Lushan (a war most people have never heard of) so it seemed to warrant a blog post, even if it is only a small part of the book. There are plenty of other stats in the first chapter – some which I take issue with – but one has to start somewhere and there are a lot of arguments in the book.

Historical Myopia and comparable violence – well yes that’s certainly true but no-one argues that there hasn’t been comparable violence. For example Niall Ferguson in ‘War of the World’ makes the argument that the Twentieth was the most violent century both numerically and in relative terms but he still spends time in the appendix pointing to other comparable violent events – e.g the Taiping Rebellion and the Mongol conquests. However the contention is that violence has declined. Now this is true, but only in the second half of the Twentieth century and principally due to the influence of nuclear weapons – an argument Pinker doesn’t properly consider (even though he has read Gaddis). Before that I suspect we were looking at a steady increase in violence – although the 20th may well have been less violent in relative terms than the 19th due to a series of destructive civil wars in China (of which the Taiping Rebellion was only the most destructive).

An Lushan open to dispute – well yes (he is basically quoting what White says on his site) but it was the figure he used in the table and having make the caveat he should have done more analysis. For example White also remarks that the figure is open to alternate interpretations but the alternate figure he quotes is still too high. The quoted rate of An Lushan is eight times the rate for the Second World War – well yes but as Baeista rightly points out it’s a totally made up number. No-one knows what the true figure is but it probably is a fraction of what he wrote and the majority of the deaths can be ascribed to disease – as with most wars before modern medicine.

David B Marshall said...

I think anyone who knows China well would have scoffed at Pinker's numbers from the get-go. More people generally die from famine and disease in conflicts like that, which can inflate the number of deaths. But the vast majority of Chinese were peasants, and even if they were military targets, for some reason, they would have been too dispersed to be easy targets. The traditional thing to do in China when trouble comes, is run away to the mountains. Even in WWII, Japanese genocide, even by airplane, was contrained by those mountains. That stretch of the Yellow River has quite a few.

I didn't know that Lushan was half Sogdian. Interestingly, so were many of the Nestorian Christians in China. The Nestorian stele makes a point of the fact that a member of the early Christian community was a big military help against the rebellion.

Anonymous said...

In those areas of the Sowietunion, which became ruled by Germany in 1941, lived 88 million people in the earliest part of 1941 - but, at least 10 million member, and maybe 15 million member of them
moved eastwards early enough to avoid the German rule. In the first half of 1942, the German army conquered newer areas ( around the Kuban river and the lower part of the course of the Don river, but,some areas, around the lower oart of the course of the Oka river ( aa bit westwards from Moscow ). At least 20 million, and maybe 26 million member of the inhbitants of the German-conquered areas dead - because the German officials similarly robbed their food, to gain empty places, which later could be settled by new settlers German people and the Northern European relatives of the German people. So,around 39 % of those Eastern-Slavic people, who were under the rule of the Nazi German government, died - they were primarily Ukrainians and Belorussians.Around 112 million people lived in the earliest part of 1941, in those areas of the Sowietunion, which never became German-ruled.

Rosie said...

Interesting & useful counter to Pinker's hugely inflated figures.

Could I hunmbly point out the it's which should be its without the apostrophe in para 3/4

Rosie said...

Useful points and interesting given Pinker's hugely inflated figures.

Could I point out the wrong use of it's in your paragraph 3 or 4 when it should be its without the apostrophe

DD said...

I would recommend reading White's entry on his cite to get both sides of the story on the interpretation of population figures of the An Lushan Rebellion. This is what he says now:

"I'll be the first to admit that the An Lushan Rebellion is problematic. Here was a civil war in China that everybody agrees was remarkably destructive. The numbers commonly cited are not wild-eyed legends passed down by frightened peasants and picked up by gullible storytellers, but rather official census records that showed a loss of 36 million people. Most scholars doubt the pinpoint
accuracy of the census, but many books still give a death toll of 36 million with no skepticism whatsoever.
As far as I could tell, there were only 3 ways to handle the An Lushan Rebellion:
1.Use the census figures as given and report an absolute population loss of 36 million.
2.Arbitrarily report a lower number based on nothing.
3.Ignore the rebellion.
Unfortunately each of those is dishonest in its own way.
Using the numbers would ignore the fact that most scholars believe the numbers to be wrong.
Making up new numbers would be pretending that I know something that no one else does.
Ignoring it would imply that it never happened.
Fortunately, the count of households presents slightly different numbers. In the seven counts before An Lushan’s Revolt, the census repeatedly found between 8 and 9 million households, and then, in the seven counts following the rebellion, the census consistently found no more than 4 million. Even a century after the revolt, in 845, the Chinese civil service could find only 4,955,151 taxpaying households, a long drop from the 9,069,154 households recorded in 755. This indicates that the actual population collapse may have been closer to one-half, or 26 million. For the sake of ranking, however, I’m being conservative and cutting this in half, counting only 13 million dead in the An Lushan Rebellion."