Ten Stems, Twelve Branches: Why Only Sixty Pairings Remain

Chujia Kuang (邝楚嘉)

Independent Researcher, Jieyang, Guangdong, China

Corresponding author: Chujia Kuang, Independent Researcher, Jieyang, Guangdong, China. E-mail: chirca@163.com

Abstract

Ten heavenly stems paired with twelve earthly branches, two at a time, ought to give one hundred and twenty arrangements. The almanac uses sixty. Every one of the sixty that is missing fails the same single rule: a yang stem pairs only with a yang branch, and a yin stem only with a yin branch. The same count of sixty follows independently from the least common multiple of ten and twelve, and from the gearing image used in teaching, in which the stems complete six turns and the branches five before the two return to 甲子 together. The round is called 一甲子, or 六十甲子, and also 花甲; one proposed derivation of that last name refers to the interleaving of odd with odd and even with even. 花甲之年 denotes sixty 虚岁, that is, nominal age, and falls close to fifty-nine completed years. The sixty are unwieldy as a single list and are cut into six sections of ten, each named after its own first member; two branches are left over in every section, and those are the 旬空. The sixty need not be memorised. What must be memorised is the six section heads.

Keywords: heavenly stems; earthly branches; yang and yin; sexagenary cycle; least common multiple; ten-day period; 花甲; 旬空

Foreign-venue shell. The abstract, keywords and reference list are added for posting outside China. The Chinese original has none of them. Everything from §1 onward is the original text, translated.

1. There is only one rule

First separate the two series by yin and yang.

甲, 丙, 戊, 庚 and 壬 are the yang stems; 乙, 丁, 己, 辛 and 癸 are the yin stems. Among the branches, 子, 寅, 辰, 午, 申 and 戌 are yang; 丑, 卯, 巳, 未, 酉 and 亥 are yin.

In pairing, a yang stem takes only a yang branch, and a yin stem only a yin branch.

甲子 holds: 甲 is a yang stem and 子 a yang branch, the two in the same column. Change it to 甲丑 and it no longer holds. 甲 is still yang, but 丑 is yin, and the two cannot be brought together.

The teaching material states this rule as 奇配奇、偶配偶, “odd pairs with odd, even with even”, and follows it with a figure: 如齿轮咬合,不可错位, “like gears in mesh, no displacement permitted”.

Of the ten stems, every one can be paired with only six branches. Five yang stems against six yang branches, and five yin stems against six yin branches; multiply the two sides and the answer is sixty.

Table of the sixty combinations: yang stems share a column only with yang branches, yin stems only with yin branches, and half of the one hundred and twenty cells stand empty

2. Another way of counting, and still sixty

The first count is the plainest. Ten times twelve is one hundred and twenty; the yin-yang rule cuts away half; sixty remain.

The second count needs one number: the least common multiple of ten and twelve, which is exactly sixty.

Take the stems and the branches as two gears with different numbers of teeth, set turning at the same moment.

The stems turn once, from 甲 to 癸, ten signs. The branches turn with them, from 子 to 酉, also ten. At this point the stems are used up, while two branches, 戌 and 亥, have not been paired off.

So the stems begin again from the top: 甲 takes 戌, 乙 takes 亥. Further on, 甲申, 乙酉, and so down the line.

The stems turn six times, sixty; the branches turn five times, also sixty. The two return to the starting point 甲子 at the same moment.

The sixty combinations complete one round, and then begin again, round upon round.

3. What this round is called

The round is called 一甲子, and also 六十甲子. Among ordinary people there is another name: 花甲.

Where 花甲 comes from has one explanation.

That explanation derives the name from 花着排, “interleaved”. In his essay 说「花甲」 (“On huajia”, Handan Daily, 28 January 2018), Li Zefu of the Handan Municipal Society for the Study of Chinese Culture gives it thus: in pairing stems with branches, an odd number takes an odd and an even takes an even, as in 甲子 and 乙丑, and never 甲丑 or 乙子; this is called 花着排, and from it the name 花甲 arises.

Apart from the name, one thing deserves separate mention: 花甲之年 means sixty 虚岁, and stands nearly a year away from the sixty years usually meant.

In the same essay Li Zefu singles this out: people who call themselves 花甲之年 and set out to celebrate a 花甲 birthday are in fact marking sixty completed years. He calls this “a misunderstanding caused by not understanding the Chinese tradition of counting years”.

Nominal age is counted from birth as one, and gains one at every lunar new year. Sixty 虚岁 therefore falls at fifty-nine completed years.

4. Sixty combinations cut into six

Sixty combinations are too long a list, so they are cut. The cut is one section per ten.

The first section runs from 甲子 to 癸酉, exactly ten days, and is called a 旬. A section takes its name from its own first member, so this one is the 甲子 旬.

Every ten thereafter forms a 旬 in turn: 甲戌旬, 甲申旬, 甲午旬, 甲辰旬, 甲寅旬. The six sections together make exactly sixty.

Positions Range
甲子旬 1–10 甲子 to 癸酉
甲戌旬 11–20 甲戌 to 癸未
甲申旬 21–30 甲申 to 癸巳
甲午旬 31–40 甲午 to 癸卯
甲辰旬 41–50 甲辰 to 癸丑
甲寅旬 51–60 甲寅 to 癸亥

One thing is hidden in that table. Every 旬 pairs ten stems against ten branches, but there are twelve branches, so two of them must go unused.

In the 甲子 旬 the two passed over are 戌 and 亥; in the 甲戌 旬 they are 申 and 酉; and so on to the 甲寅 旬, where 子 and 丑 are passed over. The two branches thus wanting are called 旬空.

How the 旬空 are taken, and what use is made of them, is another course of study. Here we stop at the name.

One word on the word: this 旬 is the same 旬 as in 一旬十天, the 旬 of 上旬, 中旬 and 下旬, the first, middle and last third of a month.

The six 旬 in a ring: the sixty combinations laid out in order, cut into sections of ten, each section named from its head

Closing

The sixty combinations laid out, and the rule is one: yang with yang, yin with yin. The two gears turn six times and five, and come back to where they started. The sixty are cut into six sections of ten days, named each from its head.

The sixty combinations need not be learned by heart. What must be learned by heart is the six section heads.

A&Q

Why is there no 甲丑?

甲 is a yang stem and 丑 a yin branch. The rule of pairing is that a yang stem takes only a yang branch and a yin stem only a yin branch; no crossing is allowed. 甲丑 breaks this rule, so it does not hold.

How are the sixty counted out?

Ten times twelve gives one hundred and twenty arrangements; the yin-yang rule removes half, and sixty remain. Or take the least common multiple: the least common multiple of ten and twelve is sixty, and the stems turn six times while the branches turn five, returning to the starting point together.

Where does the name 花甲 come from?

From 花着排, that is, in pairing stems and branches an odd takes an odd and an even takes an even, as in 甲子 and 乙丑, and never 甲丑 or 乙子. The explanation is given by Li Zefu in 说「花甲」 (Handan Daily, 28 January 2018).

Is 花甲之年 sixty completed years?

It is sixty 虚岁. Nominal age counts from birth as one and adds one at every lunar new year, so it stands nearly a year away from completed years.

What is 旬空?

The sixty combinations are one 旬 per ten, six 旬 in all. Each 旬 pairs ten stems with ten branches, and with two branches to spare, two of them must go unused. The two wanting are called 旬空. How they are taken and how they are used is another course of study.

Note by Chujia Kuang

The sixty combinations need not all be committed to memory. Learn the six section heads first: 甲子, 甲戌, 甲申, 甲午, 甲辰, 甲寅. One section a day, and in six days they are done. Once these six are held, every later exercise that comes round to a 旬 has somewhere to stand.

About the author

Chujia Kuang, courtesy name Xuzhu, is a native of Jieyang, Guangdong. He has worked on traditional Chinese calendrics since 1996 and teaches it. Since 2013 he has run classes without interruption.

This column writes about calendrics, solar terms, the stems and branches and orientation, and also about what he reads. The pen name 嘉言一得 comes from the 史记: 愚者千虑,必有一得, “the foolish man’s thousand thoughts must yield one gain”. Where something is written wrongly, corrections are welcome.

References

Kuang, Chujia 邝楚嘉. (n.d.). 四柱八字基础教材 [Foundations of Four-Pillar BaZi: teaching material]. Unpublished manuscript.

Li Zefu 栗泽甫. (2018, January 28). 说「花甲」 [On huajia]. 邯郸日报 (Handan Daily).


Author: Jiayan-Yide (Chujia Kuang 邝楚嘉) | Independent researcher of astronomical calendrics, committed to verifiable primary sources and to the open archiving of calendrical data.
Plain record · Datasets · Preprints: GitHub / WordPress / Zenodo (DOI: 10.5281/zenodo.22788686)

Three Calendars, Three Kinds of Time

Chujia Kuang (邝楚嘉)

Independent Researcher, Jieyang, Guangdong, China

Corresponding author: Chujia Kuang, Independent Researcher, Jieyang, Guangdong, China. E-mail: chirca@163.com

Abstract

In October 1582 ten calendar days were removed from the European year by decree, the better to bring the civil year back into step with the sun. That episode is used here as an entry point to the general claim that a calendar is a graduated scale imposed on the motion of the heavens. Three calendars are in ordinary use in China, and they are built on three different reference points: the tropical year, the synodic month, and the twenty-four solar terms. This note sets out the basis of each in turn, shows that the lunisolar calendar (农历, nongli) is neither a lunar nor a solar calendar but a compound of the two, and argues that the common habit of calling it the “lunar calendar” is not a loose usage but an error of classification. The stem-and-branch calendar (干支历, ganzhili) is treated as a solar calendar in its own right rather than as an appendage of the lunisolar one. A comparison table sets the four systems side by side. The note is descriptive; it contains no predictive claims.

Keywords: Chinese calendar; lunisolar calendar; tropical year; synodic month; solar terms; Gregorian reform; stem-and-branch calendar

1. A name that needs correcting

In October 1582, ten days disappeared from the European calendar. Thursday 4 October was followed directly by Friday 15 October. The intervening ten days were never lived through by anyone.

This was not a scribal error. It was an order issued by Pope Gregory XIII, and the reason was simple enough: the calendar had drifted out of step with the sun by ten days, so ten days were deleted.

The episode makes the nature of a calendar visible. A calendar is a graduated scale that human beings impose on the motion of the heavens. One revolution of the sun is taken as a year; one cycle of the moon’s phases is taken as a month. The two periods do not divide evenly into one another, and no arrangement reconciles them completely. How the reconciliation is attempted is what separates one calendar from another.

Three calendars matter in ordinary Chinese life: the solar calendar, the lunar calendar, and the stem-and-branch calendar. Their reference points are entirely different, and mixing them produces errors.

Many people call the lunisolar calendar (农历) the “lunar calendar” (阴历). The name is inaccurate, and inaccurate by a wide margin.

Calendars fall into three classes by reference point.

  • Lunar. The moon alone is observed. Months follow the phases; the sun’s position is disregarded.
  • Solar. The sun alone is observed. Years follow the tropical year; the phases are disregarded.
  • Lunisolar. Both are observed. Months follow the moon, years follow the sun, and the difference between the two is made up by intercalary months.

The Chinese lunisolar calendar belongs to the third class. It fixes the first and fifteenth of each month by the new moon and full moon, and it keeps the year length in step with the tropical year by inserting intercalary months. Observing both, it is called a lunisolar calendar. The genuinely lunar calendar is the Islamic one, which follows the moon alone.

To call the lunisolar calendar “lunar” is to describe a hybrid as though it had a single parent.

2. The solar calendar: following the sun

The reference point of a solar calendar is the tropical year, the interval between two successive passages of the sun through the vernal equinox: 365.2422 days, or 365 days 5 hours 48 minutes 46 seconds.

The difficulty is that this number is not a whole number. Counting a year as 365 days loses 0.2422 of a day each year, and in four years the loss accumulates to nearly a full day. That day has to be put back, and the putting back is the leap year.

The method has taken wrong turnings. The Julian calendar, fixed by Julius Caesar in 45 BCE, simply inserted a leap day every four years, giving an average year of 365.25 days. That exceeds the tropical year by 0.0078 of a day, or 11 minutes 14 seconds a year. The figure looks negligible, but in 128 years it amounts to a full day.

By 1582 the accumulated error had reached ten days. The vernal equinox had moved from 21 March to 11 March. The Church cared about this, because the date of Easter is computed from the equinox, and a wandering equinox meant a wandering feast. The calendar was therefore reformed, in two moves.

First, ten days were deleted: the day after 4 October 1582 was declared to be 15 October, with the days of the week continuing unbroken.

Second, the leap rule was changed: a leap day every four years, none in century years, but one again in years divisible by 400. Thus 1700, 1800 and 1900 are not leap years, and 2000 is.

Under this rule only 97 of every 400 years are leap years, giving an average year of 365.2425 days, which differs from the tropical year by 0.0003 of a day, or 25.92 seconds a year. Accumulating a full day would take more than three thousand three hundred years.

The Gregorian calendar is the civil calendar now in use.

One distinction must be kept clear: the lengths of the months in the civil calendar are arbitrary divisions and have nothing to do with the heavens. The 365 days of the year are cut into twelve parts, and which month has 31 days and which has 30 is a historical convention. February is short, on one account, because Augustus moved a day from it to August. Neither the phases of the moon nor the position of the sun enters into the division.

This is the fundamental difference between the civil calendar and the stem-and-branch calendar discussed below.

3. The lunar calendar: following the moon

The reference point of a lunar calendar is the synodic month, the interval between two successive new moons: 29.5306 days, or 29 days 12 hours 44 minutes 3 seconds.

The moon does not shine by its own light. The phases we see are the changing illuminated half of the moon as viewed from the earth. From invisible, through full, and back to invisible takes twenty-nine and a half days. The period is visible to anyone who looks up, which is why the earliest calendars generally began from the phases.

The difficulty arises here as well: twelve synodic months amount to 354.3672 days, nearly eleven days short of the tropical year.

The purely lunar calendar’s response is to leave the discrepancy alone.

The Islamic calendar works this way. Odd-numbered months have 30 days and even-numbered months 29, giving a year of 354 days. The remainder is absorbed by adding single days: eleven of every thirty years are leap years, and a leap year adds a day at the end of the twelfth month, producing a year of 355 days. It adds leap days, and never intercalary months.

The consequence is that dates drift through the seasons. The Islamic year is about 10 days 21 hours shorter than the solar year, so a month is lost every three years, and after 32.6 solar years the Islamic calendar has completed one whole additional circuit. Ramadan therefore rotates through the seasons: in some years it falls in summer, when the days are long and the fast is hard; a decade or so later it falls in winter, when the days are short.

The advantage of the purely lunar calendar is that the phases and the dates always agree. The price is that dates and seasons come apart completely. For an agrarian society dependent on the heavens, that price is not payable.

4. The stem-and-branch calendar: also solar, but divided differently

The stem-and-branch calendar is often treated as an appendage of the lunisolar calendar. It is in fact considerably more independent, and it is a purely solar calendar: it observes the sun and takes no account of the moon at all.

Its graduations are the twenty-four solar terms. The sun’s path along the ecliptic is divided into steps of 15 degrees, and 360 degrees make twenty-four terms. The sequence opens with 立春 (Beginning of Spring) and closes with 大寒 (Great Cold). The ecliptic longitudes are fixed: 立春 stands at 315 degrees, each subsequent term adds 15 degrees, and the vernal equinox returns to 0 degrees.

Months are divided by the jie (节) and not by the qi (气). Among the twenty-four terms, the odd-numbered are jie and the even-numbered are qi; one jie paired with one qi makes a month, giving twelve months in the year, with no intercalary month required. A complete revolution of the sun is a year by definition and needs no correction.

The year of the month begins at 立春, and the branch is fixed:

  • 立春 opens the 寅 month, 惊蛰 the 卯 month, 清明 the 辰 month
  • 立夏 opens the 巳 month, 芒种 the 午 month, 小暑 the 未 month
  • 立秋 opens the 申 month, 白露 the 酉 month, 寒露 the 戌 month
  • 立冬 opens the 亥 month, 大雪 the 子 month, 小寒 the 丑 month

This is the familiar 正月建寅, “the first month is established at 寅”. The 正月 here is the first month of the stem-and-branch calendar, beginning at 立春, which is not the same thing as the first day of the first month of the lunisolar calendar; that discrepancy was the subject of an earlier note in this series.

The key difference between the stem-and-branch calendar and the civil calendar lies here. Both are solar, and both follow the tropical year, but the months of the civil calendar are cut arbitrarily, whereas the months of the stem-and-branch calendar are cut by the sun’s position. February may have 28 days or 29, depending on divisibility; the 寅 month is always the stretch from 立春 to 惊蛰, depending on where the sun has reached. One follows convention; the other follows the heavens.

The counting of days by stems and branches deserves one further remark. It is the longest continuously running day-count known: from the day 己巳 in the second month of the third year of Duke Yin of Lu in the Chunqiu (720 BCE) it has run for more than two thousand seven hundred years without a break. That starting point was recovered by back-calculation from an eclipse record: the Chunqiu entry 己巳,日有食之, “on the day 己巳, the sun was eclipsed”, corresponds by astronomical computation to 22 February 720 BCE.

A complete sexagenary table is already incised on Shang oracle bone: the piece excavated at Yinxu and catalogued as 合集 37986 contains all sixty designations without omission, and served as a reference table against which dates were checked.

5. Where the lunisolar calendar stands

Return to the question of the name. The lunisolar calendar is pure on neither side, and it combines the strengths of both:

The months follow the moon. The first of the month is always the day of the new moon, and the fifteenth is always near the full moon, so that the phase can be read off the date. This is the strength of the lunar calendar.

The years follow the sun. Intercalary months restore 354 days to 365, so the seasons do not wander. This is the strength of the solar calendar.

The twenty-four solar terms are borrowed from the stem-and-branch calendar to govern agricultural timing.

The traditional rule for making up the difference is seven intercalations in nineteen years, which is approximate. Which month receives the intercalation in a given year is determined by whether that month contains a qi. The procedure is now written into the national standard GB/T 33661-2017, computed and promulgated by the Purple Mountain Observatory.

The lunisolar calendar is therefore a compound of three things: the moon governs the month, the sun governs the year, and the solar terms govern agricultural timing. To call it the lunar calendar is to name one third of it.

6. Four calendars side by side

Solar (civil) Lunar (Islamic) Stem-and-branch Lunisolar
What it observes Sun Moon Sun Moon + Sun
Year length 365 / 366 days 354 / 355 days c. 365.25 days 354 / 384 days
Month length Arbitrary, 28–31 days 29 / 30 days, by phase Two solar terms, c. 30 days 29 / 30 days, by phase
Correction Leap day Leap day, no intercalary month None required Intercalary month
Year begins 1 January 1 Muharram 立春 First day of the first month

The middle two columns are pure; the outer two are compromises.

7. What the distinction is for

The stem-and-branch calendar is used for setting out the four pillars: the day pillar and the month pillar of a birth time are determined entirely by the solar terms and have nothing to do with the first or fifteenth of a lunar month.

The civil calendar is used for daily life and for international correspondence; it counts the sun.

The lunisolar calendar is used for observing the phases and for festivals; a full moon on the fifteenth of the eighth month is something it governs.

Each of the three has its own province, and none can replace another. The consequence of mixing them is that a single date yields two results, and a dispute follows over which is right. Both are right; they are simply not the same calendar.

What comes next in this series is the solar terms themselves, where 立春 turns out not to be a whole day but a single second.

Note on the text

This is a popular exposition of material treated in the author’s teaching notes on the foundations of four-pillar practice, and it makes no claim to original research. The astronomical constants cited are standard values. This English version is issued alongside the Chinese original as the same article in another language. The author’s name is 邝楚嘉, Chujia Kuang; he writes under the pen-name 嘉言一得 (Jiayan Yide).

Declarations

Funding. This research received no external funding.

Competing interests. The author declares no competing interests.

Data availability. No new data were created or analysed in this study.

Language. The original Chinese text is deposited as a companion file to this record.


Author: Jiayan-Yide (Chujia Kuang 邝楚嘉) | Independent researcher of astronomical calendrics, committed to verifiable primary sources and to the open archiving of calendrical data.
Plain record · Datasets · Preprints: GitHub / WordPress / Zenodo (DOI: 10.5281/zenodo.22788686)

尧帝派人看星星,是为了种地

四千年前,黄河中游的一个部落首领要拍板一件事:今年什么时候下种。

早几天,一场倒春寒能把种子冻死在土里;晚几天,庄稼灌不上浆,秋霜就到了。这个决定每年都得做一次,做错的代价是一整年的口粮。

他叫尧。他给出的办法是派人去看天。

一、那十二个字,之前拆过一半

《尚书·尧典》开头记着尧的一道命令:

乃命羲和,钦若昊天,历象日月星辰,敬授民时。

这十二个字,《时间是怎么产生的?》那一篇拆过字面:「历象」是观测推算,「敬授民时」是把算出来的时节颁给百姓。

这次看另一头。这句话除了讲做什么,还交代了交给谁。开头那四个字,「乃命羲和」,是一道任命。

把报时写成人事任命,这件事本身就值得掂量。

二、四道任命,派出去四个人

《尧典》往下写,是四段几乎同构的任命。

尧派羲仲去旸谷,派羲叔去南交,派和仲去昧谷,派和叔去幽都。四个人,四个方向,各守一处观测点。要他们看什么也写明了:日中、日永、宵中、日短。

派出的人 去处 观测要点 定出的节点
羲仲 旸谷 日中 春分
羲叔 南交 日永 夏至
和仲 昧谷 宵中 秋分
和叔 幽都 日短 冬至

日中与宵中说的是昼夜等长,日永那天昼长到极点,日短那天昼短到极点。四句话,四个日子,一年的四根柱子就立住了。

南交这一处各家读法不一:《尧典》原文作「宅南交」,旧注以为下脱「曰明都」三字,另有注家读作「曰大交」,而《史记》作「居南交」。这里从经文。

四道任命与四个方向:羲仲、羲叔、和仲、和叔分赴四方,各定一个节点

四个观测点散在四方,这未必是实录,更像一种写法。但写法本身透了底:在那时的观念里,报时得有人守着、分头去看,一个人抬头瞥一眼不算数。

三、为什么值得养这么一家人

播种的窗口一年只有一次,错过就没了。报时的人手上握着的,是全族人的饭。

反过来看也成立。谁能把时节定准,谁的话就有人听。所以「敬授民时」里的那个「敬」字用得重:颁布时节这一步落定,就是一桩关乎生计的大事。

这样一份差事,交出去就不会再收回来。羲和之后,天文官成了一个世代相传的位置,旧注说羲氏、和氏世掌天地四时之官。

四、地里挖出来的那点东西

传说归传说,四千年前的人到底有没有认真看过天?考古是有话说的。

山西临汾襄汾的陶寺遗址,距今约四千三百年到三千九百年,年代正落在传说里尧舜的那一段。遗址的祭祀区有一处编号 ⅡFJT1 的建筑,十几根夯土柱圈出一道道观测缝。这套东西的用法,在《没有指南针,古人怎么定方向》里已经讲过,这里不再重复。

要在这里说清的是另一层:这些实物能证明四千年前的中原确实有人在做系统观测;它们证明不了陶寺就是尧都。

五、这条记载,能有几分信

《尚书·尧典》的成书年代远晚于尧舜。今天通行的看法是,它的定本大约成于战国,材料来源更早,但中间经过不止一次整理。

所以「尧命羲和」这四个字,未必是一份四千年前的任命书原样抄下来的。学者提醒过一种可能:后世已经有了成套的天文官制,写书的人回头往上古投射,把后来的制度安在了尧的头上。

这一层要摆在明处。可另一层也能摆在明处:即便叙述经过加工,它讲的核心并不违和。上古社会存在专职观测天象的群体,这个群体与统治权力绑在一起。考古所见与这一条不冲突,冲突只在细节。

存疑提醒:把陶寺遗址直接说成尧都,目前是假说,不是定论。这里引《尧典》,只用它说明当时的观念与制度想象,不用它给任何一个遗址定名分。

六、这套办法,今天叫什么

把一个周期的运动读出来,变成可以下发的指令,再送到每家每户手里。用今天的话说,这是一套授时系统。

它跟今天的关系比看起来近。收音机里的整点报时、手机自动校准的时间、国家标准时间的发布,做的是同一件事:把时间的读数从少数人手里,送到所有人手里。差别在精度,逻辑没换。

不过那时能定住的,只有四个点。分不出更细的段,也写不出坐标。四根柱子撑起一年,再细就没有工具了。

收尾

把报时交给一个家族,是尧那个时代给出的答案。这套答案管用了很久,久到「历象日月星辰」六个字后来成了历代治历的总纲。

但那四个人守着观测点,具体在看什么?黄昏的时候,正南方向上那颗星升到极点,凭什么能当钟用?


A&Q

问:尧典里的「乃命羲和」是什么意思?

答:「乃命羲和」是一道任命。尧把观测天象、推算时节这件事交了出去。《尚书·尧典》接下来写了四段任命,派四个人分赴四方,各守一处观测点。

问:羲和是一个人还是几个人?

答:在《尧典》的文本里,羲和是一族人的统称。往下分,是羲仲、羲叔、和仲、和叔四人,分别去旸谷、南交、昧谷、幽都四个方向。

问:那时的人凭什么定出春分、夏至、秋分、冬至?

答:《尧典》给四个人写的观测要点是日中、日永、宵中、日短,也就是昼夜等长、白昼长到极点、昼夜等长、白昼短到极点这四个日子。定住这几天,一年的四个节点就出来了。

问:陶寺遗址能证明尧的存在吗?

答:不能。陶寺的年代大致落在传说中尧舜的那一段,遗址里也确实有与观测有关的建筑,但把陶寺直接认定为尧都,目前仍是假说。

问:《尧典》的这段记载可信吗?

答:《尧典》成书远晚于尧舜,中间经过整理,还可能有把后世制度往前投射的成分。它讲的核心,也就是上古有专职观测天象的群体、并与权力绑定,与考古所见不冲突。


邝楚嘉按:这一段记载值得记的,是报时被写成了人事安排。派人、定去处、写清观测要点、交代接续的人,四件事齐了,一套观测制度才算落地。天文在这里落成了一桩政务。

相关枢纽页:历法术语表历法复原:一年从哪一天算起数据集与预印本


作者:嘉言一得(邝楚嘉)|独立天文历法爱好者,致力于可核验的原始文献与历法数据归档。
平实记录 · 数据集 · 预印本:GitHub / WordPress / Zenodo(DOI: 10.5281/zenodo.22788686)

节与气:二十四节气的两半

二十四个节气里,有十二个能换月份,另十二个不能。这一条很少被留意,却决定了节气在历法里怎么用。

2026 年秋分交节在 9 月 23 日 8 时 5 分,太阳到达黄经 180 度。它在不能换月的那一半里。

十二个月,配十二组

一年十二个月,每个月配一节一气,从头往下配:立春配雨水,惊蛰配春分,清明配谷雨,立夏配小满,芒种配夏至,小暑配大暑,立秋配处暑,白露配秋分,寒露配霜降,立冬配小雪,大雪配冬至,小寒配大寒。

每一组里,排在前的是节,排在后的是气。十二组排完,二十四段也就齐了。

各管一头

节管月份的分界。立春落进寅月,惊蛰落进卯月,一路排到小寒落进丑月。十二个节,十二个月,一处不多一处不少。

气管太阳的位置。每个气都对应黄道上的一个度数:春分是 0 度,夏至是 90 度,秋分是 180 度,冬至是 270 度。度数定下来,交节时刻就能算到秒。两件事本来不在一条线上:月份是纸面上的格子,太阳是天上走着的位置;用一节一气把它们分开标记,才不至于互相牵扯。

一个例子

秋天六段,两组交替下来:立秋是节,处暑是气,白露是节,秋分是气,寒露是节,霜降是气。

换月份换在立秋、白露、寒露。处暑、秋分、霜降只报太阳的位置,不翻月份。

所以秋分那天,太阳换了地方,纸面上的月份没有动。

气与候

还有一层差别,比节与气的分法更贴人。

太阳的位置可以算,地面上的冷暖算不准。

秋分交节在早上 8 时 5 分,可身上的凉要到寒露、霜降前后才来,差了大约一个月。

古人给这层差留了两个名目:气,指太阳走到了哪儿;,指地面冷暖到了哪一步。

刻度在先,体感在后。刻度准,体感慢。两条线对不齐,是常态。

今年的数

秋分与春分两天,正午太阳的高度角等于九十度减去当地纬度。纬度越低,太阳越高,正午的影子越短。

在北纬二十三度上下,秋分正午太阳到 66 度多;北纬三十二度上下,到 57 度多;北纬四十度上下,到 50 度。同一根八尺杆子,三地影长依次约三尺半、五尺、六尺七寸。

这几个数不必去天文台,自家院子就能量。

收束

秋分落在两个「节」中间。它到的那一天,天上的刻度先走完,地上的冷暖还没跟上。

A&Q

问:二十四节气为什么分成节与气两组?
答:一组管月份分界,一组管太阳位置。两组各十二个,交替排列。

问:秋分是节还是气?
答:是气。它落在月的当中,前后各有一个节。

问:气与候有什么分别?
答:气是太阳的位置,算得准;候是地面的冷暖,慢约一个月。


邝楚嘉按:刻度先到,体感后到。秋分就是这两条线之间那一段距离。

邝楚嘉|历法与节气

相关枢纽页:历法术语表历法复原:一年从哪一天算起数据集与预印本


作者:嘉言一得(邝楚嘉)|独立天文历法爱好者,致力于可核验的原始文献与历法数据归档。
平实记录 · 数据集 · 预印本:GitHub / WordPress / Zenodo(DOI: 10.5281/zenodo.22788686)

正月本来不是一月

先摆一个说法出来:

「正月」这两个字,本来不指一月。

它指的是岁首所在的那个月——一年的头一个月。而这个头一个月到底是哪个月,两千多年里改过很多回。

如果你穿越到秦朝过年,那个月在十月。

一、三正:哪个月当岁首

古人靠北斗七星定月份。斗柄指到哪个方位,就叫「建」哪个月,寅月、卯月、辰月是这么来的。

哪个月当一年的头,各朝不一样:

朝代 斗柄指向 岁首相当于今农历
建寅 正月
建丑 十二月
建子 十一月
建亥 十月

四朝岁首对照:横轴自十月起排,四个朝代的岁首一步步前移

这套规矩,古人叫「三正」。

胡三省注《资治通鉴》时说得最清楚:

古有三正:子为天正,周用之,以十一月为岁首;丑为地正,殷用之,以十二月为岁首;寅为人正,夏用之,以十三月(即一月)为岁首。

《史记·历书》里也有一句:

夏正以正月,殷正以十二月,周正以十一月。

上一篇讲子时的时候提过,周朝以十一月为正、以夜半为朔。说的就是这件事。

注意商和周的位置:商是十二月,周是十一月,这两个容易记反,不少文章就写颠倒了。以《史记·历书》原文为准。

二、秦朝的年,在十月

秦始皇统一六国后,用的是《颛顼历》,岁首定在建亥之月,也就是夏历十月。

所以秦朝人过年,在秋天。

有一个细节要分清:秦人不把十月叫「正月」。他们把「正月」改称「端月」,而四季的搭配照旧按夏历走——春天仍然是正、二、三月。

顺带说一句,「正月」今天读「征月」。流传最广的说法是避秦始皇的名讳,正与政同音——跟上面改「端月」是同一个由头。这个说法,清代黄生在《字诂》里已经驳过:正字本来就有平声的读法,不是避讳改出来的。

这件事后面还要用到。

汉朝初年接着用秦历,一用就是一百多年。刘邦当皇帝那阵,新年大典就在十月办。《史记·刘敬叔孙通列传》记汉高帝七年「长乐宫成,诸侯群臣皆朝十月」——长乐宫刚落成,诸侯和大臣们就赶在十月来朝贺。

三、汉武帝把年改回来了

到汉武帝太初元年,公元前 104 年,这件事定了下来。

这一年颁布《太初历》,恢复夏正建寅,重新以正月为岁首。牵头的是太史令司马迁;具体造历的二十多人里,治历官邓平和巴郡的落下闳出力最多,方士唐都分管天部。

上一篇讲十二时辰时提过《太初历》,说那是十二时辰定形的时候。这次讲的是同一部历法干的另一件事——它把岁首也定死了。

《太初历》还有个更长远的影响:它第一次把二十四节气编进历法,又立下「没有中气的月份作闰月」这条规矩。今天农历的骨架,就是那时候搭起来的。

从太初元年开始,「岁首」和「正月」合成了一件事——一年的头一个月,就是正月。

听起来平淡,但它管住了此后两千多年。

四、改回来之后,又反复了四次

改朝换代要「改正朔,易服色」,表示受命于天。所以太初历之后,岁首又动过几回:

时间 改成
公元 8—23 年 王莽 殷正(十二月为岁首)
237—239 年 魏明帝曹叡 殷正
689—700 年 武则天 周正(十一月为岁首)
761—762 年 唐肃宗 周正

中间那一段最有故事。

武则天正准备称帝。永昌元年(689 年)十一月,她下诏改用周正,把十一月定为新年的头一个月,十二月改叫「腊月」,原来的正月改叫「一月」。

后果是:永昌元年只剩下十一个月,一共 325 天。

那一年,中国人过了一个只有 325 天的年份。

她改岁首图的是政治,不是历法。

九年之后(698 年),她又动了一次手。那一年管历的人推算,转年该闰腊月;而她想要的是正月初一凑成「甲子朔旦冬至」——节气、月序、干支日序三样在同一天开头,古人眼里最硬的那种祥瑞。她挑了个由头,下令把闰月挪走,硬凑出这一天,随即改元「圣历」。

可现代学者用《麟德历》和天文工具回推,那一年的冬至其实落在壬戌日,不是甲子日。

祥瑞是造出来的。

到久视元年(700 年)十月,她又改了回去。《资治通鉴》记这一笔很干脆:「甲寅,制复以正月为十一月,一月为正月。」

五、岁首、正月、四季,是三回事

回到秦朝那个细节。

秦人以十月为岁首,却不把十月叫正月,四季也照夏历走。这说明什么?

说明在古人那里,「岁首」「正月」「四季怎么配月份」,是可以分开的三件事。

这不是后人替古人分的。唐人司马贞注《史记》,在「诸侯群臣皆朝十月」这一句下写得很清楚:

汉以十月为正,故行朝岁之礼,史家追书十月也……案:诸书并云十月为岁首,不言以十月为正月。

十月是岁首,正月还是正月。一千多年前,读史的人就把这两件事分开了。

不过,把这两件事分开,讲的还是「怎么读古书」。再往下一层,问题就变了——「三正」这套说法本身,到底站不站得住?

语言学家王力就据此认为「三正」之说并不可信——如果三代真的各用一套历法,秦朝为什么不干脆把十月改名叫正月?

这个提醒,读古书的时候很有用。

《春秋·成公元年》记了一句「二月无冰」。杜预注这一笔,只用了六个字:「周二月,今之十二月。」周历二月就是夏历十二月,正该是最冷的时候,河里没结冰,史官才要郑重记上一笔。要是按夏历二月去读,就完全读不通了——那时候早该开冻了。

不了解岁首的差别,就会以为古人记错了。

六、一个没被采纳的聪明办法

北宋的沈括想过另一条出路。

他在《梦溪笔谈·补笔谈》里说,不如干脆抛开月亮,纯用节气来排月份:

直以立春之日为孟春之一日,惊蛰为仲春之一日,大尽三十一日,小尽三十日,岁岁齐尽,永无闰余。

立春就是正月第一天,惊蛰就是二月第一天;大月三十一天,小月三十天,再也不用设闰月。

这是一部纯阳历。它跟干支历的思路几乎一样——都以节气分月,都不设闰月。前面讲三本历法时说过,干支历就是这么走的。区别在于,干支历没当成老百姓的日常历法,却一直用到今天。

沈括知道这想法会挨骂。他在这段话后头写了一句:

今此历论,尤当取怪怒攻骂,然异时必有用予之说者。

当时要被骂,将来总会有人用。九百年后读这一句,不像是对宋人说的。


两千多年里,「正月」这两个字的意思变过好几回。今天说正月初一,指的是同一件事;在秦朝说正月,指的是另一个月。

再往后,讲的是钟表上的出生时间——它不等于你出生的时辰。

相关枢纽页:历法术语表历法复原:一年从哪一天算起数据集与预印本


作者:嘉言一得(邝楚嘉)|独立天文历法爱好者,致力于可核验的原始文献与历法数据归档。
平实记录 · 数据集 · 预印本:GitHub / WordPress / Zenodo(DOI: 10.5281/zenodo.22788686)

立春不是一整天,只有一秒钟

每年 2 月上旬,朋友圈都会刷一波”今日立春”。

可要是追问一句:立春是几点几分?

十个人里九个答不上来。

答不上来很正常,因为在多数人心里,立春是一天

但它不是一天。它是一个时刻,精确到秒。

一、差十二秒,八个字换掉四个

先看一个真实的例子。

2021 年立春,交在 2 月 3 日 22 点 58 分 39 秒

这一秒的前后,落在两个不同的干支月:

  • 22:58:38 出生交节前 1 秒):庚子 己丑 壬午 辛亥
  • 22:58:50 出生(交节后 11 秒):辛丑 庚寅 壬午 辛亥

前后相差十二秒。

同一天,同一个时辰,后面两柱一模一样——年与月这两柱全换了

这不是编出来吓人的例子。每年立春、立夏、立秋、立冬前后,都要遇到。

2021 年立春 22:58:39:前后十二秒,干支年月全换

二、节气是一个”时刻”

国家标准给节气的定义很干脆:太阳地心视黄经等于 15 度整数倍的时刻

关键词是”时刻”。

黄经就是太阳在黄道上的位置。上一篇讲过,太阳一年走一圈,切出二十四个节气,每份 15 度。

所以节气的本质是:太阳走到了某个位置

那一刻跟日历翻到哪一页没有任何关系。日历是给人看的,太阳不认日历。

三、那为什么我们印象里节气是”一天”

因为公历和节气都跟着太阳走,两套刻度对着同一个东西,自然对得齐。

上半年节气多半落在 6 日和 21 日前后,下半年多半落在 8 日和 23 日前后,前后不差一两天。

对得齐不等于是一回事。日期稳定只是结果,不是原因。

真正的原因是太阳准时走到了那个位置,公历恰好也按回归年编排,两者凑巧对上了。

一旦凑不齐,日期就浮动。

四、立春有时在 3 日,有时在 4 日

交节时刻每年都在往后推。上一篇说过那个数字——每年多出几小时,攒几年就能跨过午夜零点。

看 2015 到 2030 这十六年,立春落在 2 月 3 日的只有四年:2017、2021、2025、2029

正好每隔四年一次,而且都是闰年之后的第一年——闰年二月多出一天,等于把后面所有节气的时刻往前拨了一格。

所以”立春在 2 月 4 日”这句话,十六年里要说错四回。

2021 年那次尤其悬:交在 22 点 58 分 39 秒,离当天结束只剩一小时零一分。再晚一点点,就掉到 2 月 4 日去了。

五、古人也为这件事走过弯路

最早定节气图省事:把一年按时间均分二十四份,每份 15.2 天,叫”平气”。好算,也好排日历。

问题是太阳不匀速。地球轨道是椭圆,1 月初跑得快,7 月初跑得慢,太阳在天上移动因此有快有慢。均分法假设它匀速,对不上。

隋代刘焯就提出了更准的”定气”——按太阳实际走到哪一度来定。

可唐朝人编《大衍历》时留了句小注:

凡推日月行度及轨漏交蚀,并依定气。若注历即依恒气也。

算日月食,用定气;印日历,还是用平气。

明知哪个准还用不准的,就因为平气好排日子。这个取舍一直用到明末,到清顺治二年(1645 年)才改过来——中间隔了一千零四十一年。

改用定气之后,节气之间就不等长了:冬至前后只隔十四天多,夏至前后能拉到近十六天。

六、所以,交节时刻要查到分

交节是时刻。凡按交节来定的东西,只记到日子是不够的。

遇到时刻卡在交节点前后一两小时的,还得记到分。

古书与今天的历书报交节,都报到分,有的报到秒,就是这个道理。

还有个容易混的地方:交节时刻每年都不一样。查当年的数据,国内以紫金山天文台编的日历为准,不能凭印象——同一个节气,今年和去年能差好几个钟头。


节气是时刻,不是日子。

前面讲过,民俗属相看正月初一,干支纪年看立春。这一篇补上后半句——立春本身也只是一个时刻,往前往后差几分钟,纪年就换了一年。

所以记载交节前后的事,只写日子是不够的——要追问一句:几点几分?

再往后讲,是时辰——子时不是半夜十二点,它是两个小时,而且还跨到了第二天。

相关枢纽页:历法术语表历法复原:一年从哪一天算起数据集与预印本


作者:嘉言一得(邝楚嘉)|独立天文历法爱好者,致力于可核验的原始文献与历法数据归档。
平实记录 · 数据集 · 预印本:GitHub / WordPress / Zenodo(DOI: 10.5281/zenodo.22788686)

立春换岁,还是正月初一换岁?

每年总有那么十几天,会被人反复问同一个问题:孩子到底属蛇,还是属马?

就取 2026 年春节前的那一段来说。

先看两个例子

2026 年 2 月 4 日到 2 月 16 日出生的孩子。

这一年立春在 2 月 4 日凌晨 4 点 02 分,落在农历乙巳年腊月十七;春节在 2 月 17 日。也就是说,从立春到除夕,中间隔着 13 天。

这 13 天出生的孩子:

  • 按民俗算法,年还没过,仍是乙巳蛇年腊月,属蛇
  • 按干支历算,立春已交,干支年已经是丙午

再看 2025 年 1 月 29 日到 2 月 2 日出生的孩子。

2025 年春节在 1 月 29 日,立春在 2 月 3 日,是正月初六。所以正月初一到初五这五天出生的:

  • 按民俗算法,年已经过了,是乙巳蛇年,属蛇
  • 按干支历算,立春还没到,干支年仍是甲辰

把两组摆在一起看就明白了。同样是”属蛇”,前面那组的干支年是丙午,后面那组是甲辰。民俗口中的属相没变,干支历里的年差了整整两年。

同样是属蛇,干支年差两年:2026 年立春在腊月、2025 年立春在正月初六

问题不在谁对谁错,在于它们本来就不是同一套历法

三套历法同时在跑

我们日常嘴里说的”年”,其实混着三套时间系统。

第一套是公历。 地球绕太阳一圈算一年,365 天多一点,多出来的那点靠闰年补。岁首是 1 月 1 日。这套跟干支纪年没关系,只用来记录钟表时间。

第二套是农历。 月份跟着月亮走,从这次看不到月亮到下次看不到月亮,算一个月,29 天半。12 个月加起来 354 天,比太阳年短 11 天。差的这些天,攒几年就多出一个月来,那个月就是闰月。岁首是正月初一。

第三套是干支历。 完全跟着太阳走。太阳在黄道上每走过 15 度,定一个节气,二十四个节气正好一圈。岁首是立春。

干支纪年的年与月,出自第三套

公历看太阳、农历看月亮、干支历看节气,干支纪年出自干支历

这就是为什么会出现前面那两组错位——民俗属相看的是农历的正月初一,干支纪年看的是干支历的立春,两家的”新年”根本不在同一天。

农历的正月,是从冬至倒推出来的

这一节可能有点反直觉,但很重要。

农历的编算现在是有国家标准的,2017 年发布,中国科学院紫金山天文台起草,编号 GB/T 33661-2017。里面定了五条规则,我挑三条说:

  • 月亮完全看不见的那天(朔日),定为每个月的初一
  • 冬至落在哪个月,那个月就定为农历十一月
  • 十一月往后数第二个(不数闰月),就是正月

看出来没有——农历的正月不是直接定下来的,是从冬至倒推出来的。整套历法的锚点压在冬至上,不在春节上。

闰月也是同一套逻辑。二十四节气里有一半叫”中气“,共十二个:雨水、春分、谷雨、小满、夏至、大暑、处暑、秋分、霜降、小雪、冬至、大寒。农历规定,哪个月摊不上中气,哪个月就设为闰月。

所以闰月不是凭空多出来的一个月,是那个月份确实夹在两个中气中间、够不着任何一个,才被指定为闰月。

顺便说一句:2025 年乙巳蛇年是农历闰年,384 天,年头年尾各有一个立春(2025 年 2 月 3 日和 2026 年 2 月 4 日),民俗叫”双春年”。2026 年丙午马年只有一个立春,在 2027 年 2 月 4 日,腊月廿八。再往后一年是”无春年”,一个立春都没有。这些说法指的都是在农历年内数立春的个数,跟吉凶没关系,纯粹是历法凑出来的现象。

干支历为什么用立春

回到干支历本身。

干支纪月用的是「气」,不是「月」。月相的圆缺跟气候冷暖没有对应关系,初一的月亮是圆是缺,改变不了地里的墒情。能反映寒来暑往的,是太阳在黄道上的位置。

立春那一刻,太阳到达黄经 315 度,阳气开始回升。把这一刻定为一年的起点,逻辑上是自洽的。

同样的道理,干支月也不是按农历初一换的,是按「节」换的。二十四节气里,立春、惊蛰、清明、立夏、芒种、小暑、立秋、白露、寒露、立冬、大雪、小寒这十二个叫”节”,剩下十二个叫”中气”。干支月的边界落在「节」上:立春一到就是寅月,惊蛰一到就是卯月,跟农历是不是初一没有关系。

节和中气这两班人马,各管一摊:节管分月,中气管闰月。很多人学了很多年,还把它们混为一谈。

两套口径,各管一摊

这里要说清楚一件事,免得以后在文章里被人挑毛病。

国家标准的口径是:生肖更替以农历正月初一为基准,干支纪年与农历年同步。这一点在 GB/T 33661-2017 里写得很明确,官方日历、节日、民俗活动都按这个来。

干支纪年的通行做法是:立春换年

两套口径并存,各管一摊,谁也替代不了谁。你去跟人聊”属什么”,那是民俗话题,按正月初一;你要推干支纪年,那按立春。

所以别急着下”生肖应该按立春算”这种全称判断——那是在跟国标唱反调,也容易吵起来。真要说,就说清楚:民俗属相看正月初一,干支纪年看立春。两边都不错,只是各管各的。

同一个孩子,两套口径给出两个答案:民俗上属蛇,干支年是丙午。两句话,不冲突。

还有一件事没说

上面通篇讲的是”哪一天”。但干支纪年要的是「哪一刻」。

同样一个出生日期,在乌鲁木齐出生和在揭阳出生,真太阳时能差将近两个小时。钟表上的北京时间过了一整天才换日,可太阳并没有走到那个位置。时辰交界、特别是子时交界上,这一步算错,日子就要归到前一天或后一天去。

这是后面要讲的事。

这个号写的是中国古代天文与传统历法,先从最容易被传乱的底盘讲起:三套历法的关系(本篇)、交节时刻与十二时辰的交界、真太阳时、六十甲子是怎么排出来的、闰月是怎么插进去的。

相关枢纽页:历法术语表历法复原:一年从哪一天算起数据集与预印本


作者:嘉言一得(邝楚嘉)|独立天文历法爱好者,致力于可核验的原始文献与历法数据归档。
平实记录 · 数据集 · 预印本:GitHub / WordPress / Zenodo(DOI: 10.5281/zenodo.22788686)

三本历法·三种时间

1582 年 10 月,欧洲有十天凭空不见了。

10 月 4 日星期四过完,第二天直接跳到 10 月 15 日星期五。中间那十天,从来没有人过过。

这不是传抄的笔误,是教皇格里高利十三世下的命令。原因说起来也简单:历法跟太阳对不上了,多算出十天,删掉。

一件事能看出历法的本质——它是人类给天体运行做的刻度盘。太阳走一圈算一年,月亮圆缺一轮算一月。可这两个周期除不尽,谁也凑不齐谁。怎么凑,就分出了不同的历法。

我们日常要打交道的,其实有三本:太阳历、阴历、干支历。三本的刻度基准完全不同,混着用就会出错。

一、先把一个叫法纠正过来

很多人把农历叫「阴历」。这个叫法不准确,而且差得有点远。

历法按基准分三类:

  • 阴历:只看月亮。月份跟着月相走,不管太阳到哪儿。
  • 阳历(太阳历):只看太阳。年份跟着回归年走,不管月亮圆缺。
  • 阴阳合历:两头都看。月份跟月亮,年份跟太阳,中间的差额用闰月补。

农历属于第三种。它用朔望月定初一十五,又靠闰月把年长拉回回归年——两头都占,所以叫「阴阳合历」。真正只跟月亮走的阴历是伊斯兰历,那才是纯阴历。

把农历叫阴历,等于把混血说成单亲。

三本历法对照:看谁、一年多长、一月多长、差的天怎么补

二、太阳历:跟着太阳走

太阳历的基准是回归年——太阳接连两次通过春分点的时间间隔,365.2422 天,也就是 365 天 5 小时 48 分 46 秒。

麻烦在于这个数字不是整数。一年按 365 天算,每年少 0.2422 天,四年就攒出将近一天,得补回去,这就是闰年。

补的办法走过弯路。公元前 45 年凯撒定的儒略历,简单地四年一闰,平均年长 365.25 天。比回归年多出 0.0078 天,一年差 11 分 14 秒,看着不起眼,攒 128 年就是整整一天。

到 1582 年,误差已经攒到十天。春分从 3 月 21 日提前到了 3 月 11 日。教会在意这件事——复活节的日期要按春分推算,春分飘了,节日跟着飘。于是改历,做了两件事:

一是删掉十天,把 1582 年 10 月 4 日的次日定为 10 月 15 日,星期几照旧连着算。

二是改置闰规则:四年一闰,逢整百年不闰,逢四百年再闰。所以 1700、1800、1900 都不是闰年,2000 年是。

这么一改,四百年里只剩 97 个闰年,平均年长 365.2425 天,跟回归年差 0.0003 天,一年差 25.92 秒。要攒够一整天,得等三千三百多年。

我们现在用的公历,就是这套格里高利历。

但有一件事要分清楚:公历的月份长度是人为切分的,跟天象无关。 一年 365 天切成十二份,哪个月 31 天哪个月 30 天,是历史留下的约定——二月之所以短,一种说法是奥古斯都从二月挪走一天加到了八月。月亮圆缺也好,太阳位置也好,都不参与这个划分。

这是公历和后面要说的干支历最根本的差别。

三、阴历:跟着月亮走

阴历的基准是朔望月——月亮接连两次朔(完全看不见月亮的那天)的间隔,29.5306 天,29 天 12 小时 44 分 3 秒。

月亮本身不发光,我们看到的圆缺,是它被太阳照亮的那半边在变。从看不见,到满月,再到看不见,一轮 29 天半。这个周期抬头就能看见,所以早期历法多半从月相起手。

麻烦也出在这儿:12 个朔望月是 354.3672 天,比回归年短了将近 11 天。

纯阴历的处理办法是——不处理。

伊斯兰历就是这样。单月 30 天,双月 29 天,一年 354 天。多出来的零头靠加一天解决:30 年里挑 11 年当闰年,闰年在十二月末加一天,凑成 355 天。它只加闰日,从来不设闰月。

后果是日期在季节里漂移。伊斯兰历每年比太阳年短约 10 天 21 小时,三年就短掉一个月,32.6 个太阳年之后,伊斯兰历已经整整多走了一年。斋月因此在四季里轮流转:有一年赶在夏天,白天长,斋戒难熬;再过十几年落到冬天,日子就短得多。

纯阴历的好处是月相和日期永远对得上,代价是日期和季节彻底脱钩。对靠天吃饭的农耕社会来说,这个代价付不起。

四、干支历:也是阳历,分月的办法不一样

干支历常被当成农历的附属品,其实它独立得多,而且它是一本纯阳历——只看太阳,完全不看月亮。

它的刻度是二十四节气。太阳在黄道上每走过 15 度定一个节气,走满 360 度正好一圈,二十四个。立春起头,大寒收尾。黄经度数定死的:立春是 315 度,往后每过一个节气加 15 度,到春分正好回到 0 度。

分月用「节」,不用「气」。二十四节气里单数是节、双数是气(也叫中气),一节配一气就是一个月,全年十二个月,不用设闰月——太阳走完一圈本就是一年,本来就不差。

月首从立春起算,地支固定:

  • 立春起寅月,惊蛰起卯月,清明起辰月
  • 立夏起巳月,芒种起午月,小暑起未月
  • 立秋起申月,白露起酉月,寒露起戌月
  • 立冬起亥月,大雪起子月,小寒起丑月

这就是常说的「正月建寅」。这里的「正月」是干支历的正月,从立春开始,跟农历的正月初一不是一回事,上一篇专门讲过这个错位。

干支历和公历的关键区别在这儿:两本都是阳历,年长都跟着回归年走,但公历的月是人为切的,干支历的月是太阳位置切出来的。公历的二月可以是 28 天也可以是 29 天,看年份能被几整除;干支历的寅月永远是立春到惊蛰那一段,看太阳走到哪儿。一个看约定,一个看天象。

干支纪日另有一桩值得说的事。它是目前已知连续时间最长的纪日法:从春秋鲁隐公三年(公元前 720 年)二月己巳日算起,两千七百多年,一天没断过。这个起点还是靠日食记录倒推出来的——《春秋》记的「己巳,日有食之」,天算回推正是公元前 720 年 2 月 22 日。

商代甲骨上已经刻着完整的干支表,殷墟出土的那片(合集 37986 号)六十甲子一个不少,是当时用来核对日期的谱表。

二十四节气与十二月建:一节配一气,合起来就是一个月

五、农历站在哪儿

回到开头那个叫法。农历哪一头都不纯粹,它把两头的长处凑到了一起:

月份跟月亮走,初一必是朔日,十五前后必是满月——看日子就知道月相,这是阴历的长处。

年份跟太阳走,用闰月把 354 天补回 365 天,季节不乱跑,这是阳历的长处。

二十四节气从干支历里借过来,管农时。

补足差额的老规矩叫十九年七闰,是个大致的规律;具体到哪一年闰哪个月,看这个月有没有中气。这套做法现在写进了国家标准 GB/T 33661-2017,由紫金山天文台算好了颁布。

所以农历是一本三合一的历法:月亮管月,太阳管年,节气管农时。叫它阴历,只说了三分之一。

六、四本历摆在一起看

太阳历(公历) 阴历(伊斯兰历) 干支历 农历
看谁 太阳 月亮 太阳 月亮 + 太阳
一年多长 365 / 366 天 354 / 355 天 约 365.25 天 354 / 384 天
一月多长 人为规定 28–31 天 29 / 30 天,看月相 两个节气,约 30 天 29 / 30 天,看月相
补差办法 加闰日 加闰日,不设闰月 不用补 加闰月
岁首 1 月 1 日 穆哈兰姆月 1 日 立春 正月初一

中间两列是纯粹的,两头是调和出来的。

七、分清楚有什么用

干支纪时用干支历,看那一天的年月与日子,全按节气走,跟初一十五没关系。

过日子用公历,跟国际接轨,算的是太阳。

看月相、过节日用农历,八月十五月亮圆,这个它管。

三本历各管一摊,谁也替代不了谁。混着用的结果,就是同一个日子排出两个结果,然后争论谁对——其实谁都对,只是不是同一本历。

再往后讲,就是节气:立春不是一整天,只有一秒钟。

相关枢纽页:历法术语表历法复原:一年从哪一天算起数据集与预印本


作者:嘉言一得(邝楚嘉)|独立天文历法爱好者,致力于可核验的原始文献与历法数据归档。
平实记录 · 数据集 · 预印本:GitHub / WordPress / Zenodo(DOI: 10.5281/zenodo.22788686)