American
Oriental Society, Western Branch
1994 Annual Meeting, Portland, Oregon
21-23 October 1994
David S. Nivison
Stanford University, emeritus
When I was in graduate school in Harvard University almost
fifty years ago, my teachers taught me that the so-called "modern
text" of the Bamboo Annals (Jin ben Zhu shu ji nian) is a faked
text, probably done in the Ming Dynasty. The faker was said to
have pretended his work to be the supposedly lost chronicle that
had been buried in 299 (or 296) BC, and discovered around 280 AD.
He had used a few historical facts but had invented freely,
adding purely imaginary dates for all events and reigns. The
precious original had disappeared centuries earlier.
Modern scholars have collected quotations from the (real)
text found in historical commentaries and encyclopedias, calling
them the "ancient text," i.e. the genuine text, or what there is
left of it. The greatest scholars in China in the 18th century
had assured us of the fraudulence of the "modern text," and the
modern scholar Wang Guowei (and others before him) had "proved"
it, by painstakingly finding the source of every item in it (or
almost: he did not account for the dates). I took all this for
granted until 1979, when I accidentally found that I could use
the Annals to help solve dating problems in the texts of bronze
inscriptions of Western Zhou, mid-11th to early 8th centuries BC.
E. Shaughnessy, in an article in 1986, has shown that parts of
the book are, verbatim, the original text. This is probably true
of almost all of it.
I have spent much time in the past 15 years pursuing this
discovery, making many mistakes and soon correcting some of them.
Perhaps what I think I discovered in July of this year is really
another mistake. You be the judges.
Very simply, I am putting together two "results" -- as I
would call them. (Others have sometimes been less kind.) I will
show that together they imply something that one would not expect
to be able to know, that if true is very unlikely to be
accidentally true. And then, by a completely independent
argument, I will show that what they imply is indeed true.
The first "result": In 1990, in Early China 15, I published
an article together with K. D. Pang, arguing that the eclipse of
the sun traditionally assigned to the reign of the fourth Xia
king Zhong Kang, and in the Annals to his fifth year, was the
eclipse of 16 October 1876 BC; and further, that this is the year
implied by a simple historical calculation: One starts with the
conjunction of planets in February of 1953 BC, which Pankenier
has argued marked the first year of effective power of Yu, the
first ruler of Xia. One then provisionally accepts the reign
lengths of the first four Xia kings in the Annals, and supposes
that between each reign there was a two-year interregnum for
completion of mourning. (The Annals has interregnums between Xia
reigns, and about a third of them are exactly two years, which
would be the expected number.) It then turns out that the fifth
year of Zhong Kang is 1876, and the stated month date converts to
16 October. And the eclipse occurred when the sun was exactly in
the spot in the zodiac that tradition said it was in.
Also in 1990, I presented a still unpublished paper in Los
Angeles in which I extended this argument through to the end of
the Xia Dynasty, taking the reign lengths given in the Annals and
assuming two years after each death of a ruler. What I ought to
get, as the last year, was 1555 BC. Pankenier had already
established that, and I had more reasons of my own. What I found
was that it was the next-to-last king, Fa, whose reign ended in
1555. I then reviewed accounts in the Annals and elsewhere of
the supposed last king Jie, called "Di Gui" in the Annals, and
found multiple reasons for concluding that Jie was a literary
invention, probably a product of the philosophical imagination of
the middle or late 400's BC. ("Di Gui" may actually have been
another name for the real last king, Fa.)
This much gives me exact dates of all the Xia kings. I have
supplied you with the complete list, even though I am concerned
here only with a few kings at the end.
Next, I went to work on the following dynasty, the Shang.
The Annals account gives reign lengths, but no interregnums. The
name of each Shang king ends in a syllable that is the name of
one of the days in the age-old Chinese ten-day week. Much work
has been done attempting to explain these "gan" names. I had
noticed years ago that the date I had derived for the first year
of the next-to-last Shang king Di Yi, 1105 BC, began with a lunar
month beginning with an yi day in the sixty-day cycle. But when
I had tried this out on other kings, it didn't work. Only last
year did I see that probably the chronology for the Shang in the
Annals was like that for the Western Zhou: There, there were in
effect two-year breaks between royal calendars, but sometime
around 400 BC people forgot about them, so that the whole
chronology shrank, towards the ends and away from the middle.
The result had been that the king in the middle, Mu Wang, who
reigned a quite respectable 39 years anyway, had his reign
bloated to 55 years, the way it is in the Shiji and in the
Annals.
The Shang list too has a king in the middle with a very long
reign, Da Wu, with an impossible 75 years. (We are talking about
something that got established fairly early on, becuase you find
this 75 years already in the "Wu Yi" chaper of the Shang shu.)
Quick experimentation showed me that two year gaps didn't work.
But three years did: The last ruler, Di Xin, had a calendar
beginning with 1086 BC, as many inscriptions (combined with
astronomy, and with the Bamboo Annals) show. 1086 does not begin
with a xin day, but 1089 does. Inscriptions, and the date for Di
Yi (and again the Annals) show that the reign calendar preceding
Di Yi, i.e., Wenwu Ding, began in 1118 BC. 1118 did not begin
with a ding day, but 1121 did.
Encouraged, I went back to the beginning of Shang. There
was funny-business after the death of the founder, Tang (a.k.a Da
Yi), which occurred (I have reasoned) in early 1542. His prime
minister Yi Yin, taking over during mourning, was manouvering
toward the throne himself. He exiled Tang's grandson and
successor Da Jia, and put two puppets on the throne by turns,
Zhong Ren and Wai Bing. But I found a plausible sequence of
dates that explained them all, and I have supplied you with this
too. The scheme posits three-year breaks between calendars, with
the first day of the succession year giving the king's gan unless
that gan would be the gan of the predecessor, in which case the
proper calendar first year is used instead. I carry the scheme
down to that long-lasting king Da Wu. I think he reigned much
less than 75 years. But more interesting, a 12-year reign of a
"Yong Ji" is in all the received chronologies placed before Da
Wu, whereas oracle inscriptions prove that Yong Ji (or Lu Ji)
belongs after him. Why?
Da Wu of Shang is like Mu Wang of Zhou not just in that
theirs are the two longest reigns in the received accounts of
their dynasties, but also in two other respects. Both are kings
in the fifth generation starting with the founder. And both have
reigns that are supposed to begin exactly 100 years after the
founder proclaimed his imperium. In Da Wu's case, the year is
1475, just 100 years after 1575, which in the Annals is the first
year in Tang's year count, before his conquest of Xia. This
wasn't his succession year but his proclamation year; and we know
this, and know furthermore that it is right, because both
dynasties, as Pankenier has shown. were heralded by signs from
Heaven. For the Shang, this was a dramatic formation of the
planets at the end of 1576, described in the Annals (where it is
misdated to 1580), and identified by Pankenier -- his most
important discovery.
So as the Shang chronology was perfected by the historians
who worked on it, 1475 must have been a fixed date. But 3-year
mourning intervals for four reigns had disappeared, and something
had to fill the resulting 12-year gap. So (I am assuming) Da
Wu's successor was given a 12-year reign, and was thought
actually to have preceded him. Again I will not recite a list of
names and figures, and will ask you to examine the work I have
put in your hands.
I would be the last to insist now that you have no choice
but to believe all this. I don't even dare to put before you my
working out of the rest of the Shang king list, which is much
more problematic. But I do think that there is enough here to be
worth pursuing; and this takes me back to late Xia.
The fourteenth Xia king Kong Jia has a short but colorful
account in the Annals. He is said, for example, to have had an
officer named Liu Lei, who was a trainer of dragons, delighting
the king. A female dragon in Liu's care died; Liu pickled it, and
fed the pickle to the unsuspecting Kong Jia, who liked the
flavor, but then later asked where the dragon had gone. So Liu
had to flee.
But this is not why Kong Jia is important to me. Kong Jia
is the only Xia king whose regularly used name ends with a gan.
"Kong Jia" means, literally, "great jia," jia being the first of
the ten day-names. Why "great"? Just possibly because what the
name picks out is not just a jia-day, but the first jia-day of
the sixty-day cycle, jiazi. But to go on, we need to think we
know, independently, what the first day of Kong Jia's reign was.
This is why my first "result" is necessary. It tells me that
Kong Jia's predecessor died in 1580, so that Kong Jia himself
succeeded in 1579, and had a calendar that began in 1577. He was
a Xia king, and the kind of lunar calendar that is supposed to
have been used in the Xia era took the lunar month preceding the
month containing the spring equinox as the first month of the
year. So what I have to identify is the first day -- the "dark
of-the-moon" day or "syzygy" day, I will assume -- of that month,
in a Chinese year more or less coinciding with the year 1577 BC.
Here is one way of doing it, if you don't have a computer
program that does it for you. The relation of first days of
lunar months to days in the solar year is repeated approximately
every 19 years. Take forty 19's, = 760, and consult Herman H.
Goldstine's New and Full Moons, 1000 B.C. to A.D. 1651
(Philadelphia: American Philosophical Society, 1973), for the
year 1577 - 760, = 817 BC. The new moon beginning the pre-spring-
equinox month in that year occurred on 15 February. The 19-year
cycle if used without correction would yield a date that would be
one day too early for every 310 years back, so 15 February should
be two days early. Next, consult W. D. Stahlman and Owen
Gingerich, Solar and Planetary Longitudes for Years -2500 to
+2000 by Ten-day Intervals (Madison: University of Wisconsin
Press, 1963), for the year 1577 BC (i.e., -1576), for 17
February. This day had Julian Day number 114 5471. To obtain
the 60-day cycle number for that day, divide the JD number by 60
and subtract 10 from the remainder. The answer in this case is
day one in the cycle, jiazi.
Finally, go to Paul Ahnert's Astronomisch-Chronologische
Tafeln fur Sonne, Mond und Planeten (Leipzig, 1960), and check
out 17 February 1577 BC to see if it was in fact a first-of-month
day. One finds that it was. I have given you the calculation,
face-up.
What does one make of this? Certainty we do not have; but
then, we never have it. There is a one-in-sixty chance that a
day selected blind might turn out to be a jiazi day. Looking
again at the Bamboo Annals, one finds that Kong Jia's predecessor
king Jin had another name sometimes used, "Yin Jia." The most
obvious meaning is "succession-jia." So, look again at Nivison's
table of dates for Xia kings, finding that Jin's succession year
(the year following the year of death of his predecessor) was
1589 BC; and now go through the same calculation, finding that
this year in the Xia calendar probably began with day jiaxu.
There was a one-in-ten chance that this date might accidentally
turn out to be a jia day; and so altogether a 1-in-600 chance
that the two jia names would turn out accidentally to correspond
to expected jia days. Not too bad.
Shall we say, then -- cautiously -- that the Shang
institution of gan names for kings was being anticipated by Xia
rulers? Perhaps not: First, we deal with only two kings, #13
and #14 out of a list of sixteen. Second, they are successive
kings, yet they have the same gan; and the Shang would never do
that. Third, there are, in the Annals, a few pre-dynastic Shang
dates and names to check; and when I check them, I find the Shang
convention already in use during the Xia era. Explaining this
would take time that I don't have; but I have supplied you with
these data too.
So I see the matter this way: The gan name convention for
rulers was used by the Shang royal ancestors for centuries before
the Shang dynasty. Toward the end of the Xia era, it began to be
used by others too, at least by Xia kings, but incompletely and
only occasionally.
But, of course, the real news is that something all of us
learned in school -- and something some of us still vehemently
insist -- isn't right: Xia has been "mythicized," but Xia isn't
a myth. There really was a Xia Dynasty. And I seem to have the
names and exact dates of its kings, every one of them. What the
Xia was -- a sophisticated urban civilization, with writing; or
just a collection of dirty stone-age villages with common chiefs
who somehow got their names and years remembered; or something in
between -- this is a different question entirely, for which I
have no answer.
And there is one more thing to recognize from this inquiry:
The "modern text" Bamboo Annals -- called "modern" only
apologetically, because it is supposed to be a fake -- is turning
out to be a veritable rosetta stone for figuring out the exact
dating of unbelievably early Chinese history.
These several matters of fact to which my argument points
are important, I think. But before I close, I want to redirect
your attention to the form of my argument itself. I do this
because it has been my experience that most historians dismiss
this kind of argument as unacceptably speculative.
I began with two hypotheses that I thought I had established
reaonably well (by argumentation similar to the present argument,
actually). One gave me the exact dates of the Xia kings; the
other explained the gan names of the Shang kings as being derived
usually from the first days of their reigns. I put these two
hypotheses together into a single theory, and found a test case,
in the late Xia king Kong Jia, conspicuous because he, unlike
other Xia kings, had a gan name. The first hypothesis gave me
his exact dates. The second one told me that the first day of
his reign, given those dates, ought to be a jia day, and probably
a jiazi day. Then by a completely independent calculation, owing
nothing to history, I found that the indicated day was a jiazi
day. I argue now that the best explanation of this "predictive"
success is that each of my two hypotheses, as a whole, is true.