《説郛》版本史 ("Textual History of Tao Zongyi's Shuofu")

馬可·波羅研究

《説郛》版本史

——《聖武親征録》版本譜系研究的初步成果①

艾騖德（Christopher P. Atwood）撰

馬曉林譯

引言

研究宋元文獻者早就對《説郛》很熟悉了。這部大型類書是陶宗儀（字九

成，號南村，1316—1403）②在元明之際亂世中編纂的，最初編成於 1361 年③。

作者單位：美國印第安納大學

譯者單位：南開大學歷史學院

① 我對《聖武親征録》和《説郛》的研究得到了許多同行的熱情幫助。我要特别感謝賈晉珠教授

（Lucille Chia, University of California Davis)、党寶海教授（北京大學）、俞小明女士（臺北“中央圖書

館”）、劉雯玲女士（Indiana University, Herman B. Wells Library）、松田孝一教授（大阪國際大學）、中見

立夫教授（東京外國語大學）、橘誠（下關市立大學)、齊光教授（復旦大學）、王函女士（中國國家圖書

館）、吴志坚博士（浙江圖書館）、徐三見先生（臨海市博物館）、周嶠先生（復旦大學）、周卿先生（上海圖

書館）。我也要特别感謝馬曉林的認真閲讀，使我注意到本文初稿的幾處錯誤。若本文在史實或闡釋

上仍有失誤，當然皆由我個人負責。

② 陶宗儀生卒年仍有争議，兹從昌彼得之説。昌彼得，《説郛考》，首版 1962；修訂增補版，臺北：

文史哲出版社，1979，407—482頁。

③ 直到最近，只能説《説郛》最早的版本比成書於 1366年的陶宗儀的名著《南村輟耕録》要早（昌

彼得，《説郛考》，12—13頁）。不過，毛氏汲古閣抄本保存了最早的 1361年稿本；參看徐三見，《汲古閣

藏明抄六十卷本〈説郛〉考述》，《東南文化》1994年第6期，112頁。 397

國際漢學研究通訊

類書作爲處理大幅增多的文獻的一種方法，自宋代已頗普及①。《説郛》是類書

的一個範例。入明，陶宗儀繼續增補《説郛》。在殘存的忠元心理驅使下，他

編選的作品中含有大量的内陸亞洲主題②。

陶宗儀常常被批評輕信和迷信，但他對異域的開放心態，以及在元朝治

下的生活，似乎使他比其他的中國學者更爲留意更廣闊的世界。他 1366年撰

成的彙集了奇聞軼事、研究筆記、老生常談的《南村輟耕録》一書，含有大量關

於元朝蒙古人、色目人及其他人群的信息③。陶宗儀所引録的稀見而隱秘的

資料中，有蒙古皇族的官方世系“十祖世系録”④。《説郛》中，則既有《聖武親征

録》——成吉思汗、窩闊台《實録》的簡編本，也有《蒙韃備録》——唯一一部在

成吉思汗在世時寫成的關於蒙古的史料⑤。在 1376年出版的《書史會要》中，

陶宗儀在諸朝代的章節中收録了八思巴字、契丹字，又在“外域”一節中收録

了畏兀兒、天竺、日本國、回回（阿拉伯）文字⑥。在《説郛》中，他收録了很多關

於内亞王朝、東南亞王國的史料。若没有他的收録，這些史料可能就失傳了

① 關於類書，見 EndymionWilkinson, Chinese History: A Manual, Cambridge& London:Harvard

University Press, 1998, pp. 601—612, 以及 Chinese History: A New Manual, Cambridge& London:

Harvard University Press, 2013, pp. 955—962.

② 關於陶宗儀的生平，見昌彼得，《説郛考》，2—10 頁，407—482 頁；以及牟復禮（Frederick W.

Mote），T’ao Tsung-i and His Cho Keng Lu (Ph.D. Dissertation, University of Washington, 1954), pp. 1—

12, 15—77, 壓縮版見其文章：“Notes on the Life of T’ao Tsung-i”, Silver Jubilee Volume of the ZinbunKagaku Kenkyusyo, Kyoto University (Kyoto: Research Institute in Humanities of Kyoto University,

1954), pp. 279—293。孫作 1374年撰寫的陶宗儀小傳是惟一的主要史料，英譯見 Mote, T’ao Tsung-i,pp. 29—31。昌彼得彙集許多零散材料，描繪出一幅更全面的圖景。

③ 見牟復禮的臚列。Mote, T’ao Zongyi, pp. 147, 149——150, 160.

④《元史》卷一〇七，2729 頁；參看伯希和注，載 L. Hambis, Le chapitre CVIIde Yuan che, Leiden:

E. J. Brill, 1945, p. 144，以及 Christopher P. Atwood,“Six Pre-Chinggisid Genealogies in the Mongol

Empire”, Archivum Eurasiae Medii Aevi, vol. 19 (2012), pp. 5—57.

⑤ 研究見王國維，《蒙韃備録箋證》，載《蒙古史料四種》，1926，臺北：正中書局，1962重印，431—

457 頁 (1a-14b) + 459—464 頁。各種語言譯本有：Nikolai Tsyrendorzhievich Munkuev, trans., Men-dabei-lu:“Polnoe opisanie Mongolo-Tatar”: faksimile ksilografa (Moskva, 1975); Peter Olbricht und

Elisabeth Pinks, trans., based on the draft of Erich Haenisch und Yao Ts`ung-wu, Meng-Ta pei-lu undHei-Ta shih-lüeh: chinesische Gesandtenberichte über die frühen Mongolen 1221 und 1237 (Wiesbaden,

1980); Möngkejayag-a, trans., in Ge. Asaraltu and Köke’öndör, ed., Bogda Bagatur bey-e-ber dayilagsantemdeglel (Höhhot, 1985), pp. 93—156.

⑥ Mote, T’ao Zongyi, pp. 82—87, esp. 85-86.陶宗儀、朱謀垔著，徐美潔點校，《書史會要·續書史

會要》，杭州：浙江人民美術出版社，2012，第8卷，第234—39、225—26、229頁。398


（見表1）。

幾乎整個明代，《説郛》只是以抄本形式流傳。大概 250年後，即崇禎年間

（1628—1644）才被杭州宛委山堂雕版付印。刻本剔除了《説郛》的一些内容

（包括《聖武親征録》在内），加入了一些新内容，多數明代抄本爲 100卷，刻本

則擴增至 120卷①。清初，這批雕版被多次利用，不僅内容被重新編排，且率先

删除了出版者認爲可能觸犯當朝滿族統治者的含有評論“蠻夷”内容的宋代

作品——如《蒙韃備録》便是這次清理的犧牲品。

迄今的《説郛》研究

《説郛》收録了很多獨家的重要史料，利用它的最大障礙就是版本上的難

題。抄本和刻本都是以篇幅和結構完全不同的很多版本流傳於世。哪個版

本最早，各版本之間相互關係如何，對於這些問題，著名學者如昌彼得、饒宗

頤給出的答案非常不同。同時，中國大陸目録學的發展，使越來越多的明代

抄本得到辨識并編目。這些抄本在《説郛》版本史上是什麽位置，仍不清楚。

對清代學界來説，《説郛》是一部 120卷的書，内容主要聚焦於經典與文學

主題——明代抄本以及早期 120卷刻本中很有特色的關於内亞、海外異國的

内容大部分被删除了。不過，書目偶爾記載存世的各種《説郛》具有不同的長

度，大多是 100卷，也有 60或 70卷。學者如王國維、伯希和讀了《説郛》抄本後

發現，《真臘風土記》《聖武親征録》《蒙韃備録》等作品的《説郛》抄本遠遠優於

其他存世抄本（實際上，這些存世抄本皆出自《説郛》，但當時的學者不一定清

楚這一點），遂産生了濃厚興趣。

現代《説郛》研究之始便是他們在 1920年代的工作，關注明末清初諸刻本

之間的關係及它們與郁文博 1496年明中期本的關係。郁文博序見於《説郛》

初刻本，通常認爲（實際上並非如此）他的編輯活動肯定對抄本系統很重要。

① 關於《説郛》的這個刻本有很多文獻，但是仍然存在許多問題。見昌彼得，《説郛考》。《説郛三

種》（上海：上海古籍出版社，1988）重印了崇禎刻本的全部内容，但是令人混淆的是加入了李際期序與

王應昌序，它們是在清代 1646年重印時才首次加入的。同樣，《東方文化學院京都研究所漢籍目録》

（京都：東方文化學院京都研究所，1938）324—347 頁述其所藏《説郛》，令人混淆地繫於順治三年

（1646）、李際期贊助，實際上它是明代印本。 399


這是昌彼得《説郛考》基於倉田淳之助、景培元、渡邊幸三、伯希和等人的先行

研究而得出的結論①。

《説郛》研究的里程碑是 1927 年上海商務印書館排印出版 100 卷本《説

郛》。張宗祥編輯這一版本，試圖回溯至 120卷以前，重建 1620年代以前的抄

本形態②。爲此，張宗祥利用了四種殘抄本，皆爲 100卷本但都不全，其中三種

收《聖武親征録》於第 55卷。第一種爲傅增湘所得《説郛》抄本殘卷，王國維以

之建立《聖武親征録》説郛本，其中部分内容成於15世紀末、1505年③。第二種

爲上海商務印書館涵芬樓藏萬曆（1573—1619）抄本④。第三種是孫詒讓藏

本，張宗祥複製了一部⑤。令《説郛》研究者們很失望的是張宗祥校勘抄本時

没有遵守學術規範，張宗祥重建陶宗儀《説郛》原本之説也成爲學者們攻擊的

目標。渡邊幸三及其他學者早就指出永樂（1403—1424）本的存在，駁斥了張

宗祥重建洪武年間（1368—1399）陶宗儀原本之説⑥。事實上，張宗祥所用的

所有抄本都是 1450年以後的本子，其中一些内容只可能是陶宗儀死後才加進

去的。不過，儘管如此，這個新版本也遠比120卷刊本更接近陶宗儀的原本。

昌彼得的書出版以後，中國以外的研究陷入停滯。中國大陸以外只存在

兩種明代抄本《説郛》（一在香港，一在臺灣），所以直到 1970年代末中國大陸

的學術活動恢復以後，《説郛》研究才慢慢重新開始。鑑於張宗祥本大概足以

代表 725目、100卷的抄本系統，研究便轉而關注辨識并著録獨立於 100卷抄

本系統的《説郛》抄本，以及諸 100卷抄本與陶宗儀原本的關係。《説郛》抄本數

① 倉田淳之助，《「説郛」版本諸説と私見》，Silver Jubilee Volume of the Jinbun Kagaku Kenkyūsyo,Kyoto University, Kyoto: Kyoto University Humanities Scholarship Research Center, 1950, pp. 287—304；

King P’ei-yuan 景培元，Études comparative des diverses éditions du Chouo fou, Scripta Sinica Monograph

series ，Beiping: Centre Franco-Chinois d’Études Sinologiques, 1946；渡邊幸三，《説郛考》，《東方學報》

京都 9（1938），218—260頁；Paul Pelliot,“Quelques remarques sur le Chouo fou”, T’oung Pao 23 (1924),

pp. 163—220.

② 陶宗儀，《説郛》，上海：商務印書舘，1927，一百卷，40册，涵芬樓明抄本。涵芬樓一百卷本。臺

北商務印書舘 1972年重印，1988年重印載於《説郛三種》十册的前兩册。後者第十册有書目索引。昌

彼得，《説郛考》，43—405頁，483—506頁提供了張宗祥版本中的全部書目的索引及敘録。

③ 兹從賈敬顔，將此混合本名爲傅本（Fu），詳見附録。

④ 兹從賈敬顔，將此本名爲張本（Zg），詳見附録。

⑤ 我命名此本爲孫本（Sn），詳見附録。

⑥ 渡邊幸三，《説郛考》，230頁；King P’ei-Yuan, Etude comparative, pp. 3—4.400


量龐大，各抄本基本上都有缺卷，所以學者們主要注重嘗試以明代學術著作

——其中提到有更古的、非 100卷本《説郛》的存在——中的那些隱晦的提示

來匹配《説郛》目録。

譜系法（Stemmatic）之研究趨向

與此同時，北京中央民族大學歷史系教授賈敬顔（字伯顔，1924—1990）

開拓了一個完全不同的研究趨向。他的《聖武親征録》注釋本於 1979年完成

并油印，其中他鑑定了幾種包含《聖武親征録》的《説郛》抄本，通過細緻的文

本比較，將這些抄本由劣到優進行了試排序。他又將這幾個抄本加上簡稱并

予著録，總結了在中國稀見抄本的目録。如果能夠廣爲流傳，這部著作對《説

郛》研究的意義會是顯而易見，但因爲只是油印本，賈敬顔本《聖武親征録》在

蒙古學研究者之外没有得到應有的承認和廣泛的流傳。

賈敬顔的研究表明，研究《説郛》諸本的關係，不僅可以對其内容目録與

作品編排做大的考察，還可以對《説郛》中的某一個（或多個）作品進行細緻的

研究。用校勘學（textual criticism）的譜系法（stemmatic methodology），考察以

抄本形式複製的作品，可以確定一個抄本如何從它的一個或多個範本

（exemplar）中繼承特定的標誌性訛誤（indicative errors），然後又加入一些這樣

的訛誤，傳給從這裏複製的所有抄本，如此等等。因此，對多種抄本的細緻考

察可以讓研究者畫出一個譜系（stemma），顯示出各抄本之間的關係。儘管賈

敬顔没有使用譜系法，但他找出并初步著録了現存於各圖書館的大部分相關

抄本，爲使用譜系法開拓了路徑。

要畫這樣一個譜系，實踐起來最大的困難是判定哪個異寫是原初的

（primitive），哪個異寫是派生的（derived）。特别是兩種都能讀得通的時候，這

種判定常常主觀得令人沮喪。這時，故國元朝帶給陶宗儀的對異域的興趣提

供了重要的幫助。《聖武親征録》尤其適於這種分析，因爲它既有完整的波斯

文平行文獻，也有一部分蒙古文平行文獻。而且，大量的蒙古文轉寫導致漢

文抄本中的訛誤通常可以立刻被察覺，因爲它們不能還原爲蒙古文。

这种做法的效用可以用兩個例子來證明：

一個人名，一些抄本皆作“盃禄可汗”，第一個字“盃”另有抄本作“杯” 401


“孟”“盈”“覔”。對照蒙古文史書，這個名字對應的是 Buyruq Qa’an，那麽

“盃”大概是原初的寫形，而“孟”“盈”“覔”是不對的，這些寫形定然是文本的

錯訛（textual corruption）。

第二个例子是癸酉年秋蒙古攻涿州之事。有些抄本作“刻日”，其他抄本

作“二命日”、“二卜命日”、“二十餘日”。一些寫法可能比另一些更易接受，但

很難做出判定，幸好有波斯文史料。拉施特《史集》明言，蒙古軍隊“圍城二十

天後，將城攻了下來”①。因此，毫無疑問最后一種寫法是原初的，其他的或多

或少都有訛誤。《蒙古秘史》是《聖武親征録》的主要史源，其中的平行記載也

有助於判定哪个是原初的哪个是派生的。將這些證據集合起来，就能製成一

個清晰的譜系，顯示出《説郛》諸抄本中《聖武親征録》諸文本之間的關係。這

一譜系可以慎重地作爲《説郛》抄本譜系的基本假設。

像賈敬顔、王國維、伯希和一樣，我對於《説郛》版本的興趣是我研究《聖

武親征録》的成果之一。爲了準備《聖武親征録》校注本，我收集了九種《説

郛》版本中的《聖武親征録》文本，也考察了前人所提供而今不知所在的三種

《説郛》版本中的《聖武親征録》文本。這九種幾乎涵蓋了中國善本書目録所

列及（或）相關文獻提及的所有主要《説郛》抄本，僅有兩種《説郛》抄本不在其

中。名稱（採用賈敬顔之簡稱）及現藏地如下，按原初至派生順序排列：

趙本：國家圖書館，北京

鈕本：國家圖書館，北京

孫本：玉海樓，浙江瑞安

傅3本②：上海圖書館，上海

臺本：中央圖書館，臺北

汪本：浙江圖書館，杭州

① Rashid ad-Din, trans. O.I. Smirnova, notes by B.I. Pankratov and O.I. Smirnova, ed. A.A.

Semenov, Sbornik letopisei, Vol. I, part 2，Moscow: Academy of Sciences Press, 1952, p. 169;

Rashiduddin Fazlullah, trans. and ed. W.M. Thackston, Jami‘u’t-Tawarikh: Compendium of Chronicles: AHistory of the Mongols，Cambridge: Harvard University Press, 1998, vol. 1, p. 219.譯者按：漢譯文據余大

鈞、周建奇譯，《史集》第一卷第二分册，北京：商務印書館，1985，233頁。

② 傅本是由三四個不同的《説郛》殘本配在一起組成的一部近乎完整的《説郛》。《聖武親征録》在

第三部分，所以我簡稱之爲傅3本（Fu3）。402


史本：國家圖書館，北京

沈本：馮平山圖書館，香港大學

前人做過校訂而今不知所在的三種：

閶本：蘇州。鄭傑1778年《聖武親征録》研究稿本中所用。

涉本：王國維 1926年《聖武親征録》本所用。王國維注亦爲賈敬顔本

所用。

傅—京（師）本：張宗祥 1927年準備商務印書館版《説郛》時使用的傅

本抄本，賈敬顔《聖武親征録》亦用之。因其爲傅本之一種，使用“京師圖

書館鈔”書紙而名之。

關於抄本的更多信息見附録。

我的譜系法研究得出了一些重要結論，例如，香港的沈本並非如其序和

饒宗頤文中所言接近陶宗儀原本，而實際上是一個較爲晚出且錯訛較多的明

代中期版本；昌彼得所據的臺北抄本實際上是20世紀僞造的。

我也有機會考察了另外兩種不含《聖武親征録》的抄本，即臨海市博物館

藏毛氏汲古閣抄本、國家圖書館藏滹南書舍抄本。此二抄本不含《聖武親征

録》，因此不能直接加入譜系中，它們的編排非常有特色，我相信它們可以暫

與《説郛》文本演變圖相聯。

根據篇幅與（或）結構，現存《説郛》諸本（包括最初的刻本）可以分成五個

版本系統（recension）。另外還有一個版本系統，惜今不存世，但其基本的編排

可以從另外兩個系統的内證中總結出來。按照校勘學的通常做法，我將它們

標以希臘字母，列出最早的範本完成的確切或大約年代、現存範本，如下：

α系統：60卷，366目；1361年；現存毛本

β系統：100 卷，約 600 目；約 1370 年；今不存，但可用γ大致還原其

内容

γ系統：100卷，725目；約 1440年；現存趙本、鈕本、孫本、張本、傅本、

臺本、汪本、1927年商務印書館版

δ系統：100卷，662目；1496年？；現存滹本

ε系統：69 卷（無卷號），估計 725 目；嘉靖年間（1521-1566）？；現存史

本、沈本 403


ζ系統：120卷；1236目；約1615年；現存宛委山堂刻本

本文將考察這些版本系統的情况，它們如何産生，以及它們的相互關係。

α系統

據陶宗儀的友人楊維楨在序中所述，《説郛》最初有60卷：

天台陶君九成取經史傳記下迨百氏雜説之書千餘家，纂成六十卷，

凡數萬條，剪揚子語名之曰《説郛》，①徵予序引。閲之經月，能補予考索

之遺。學者得是書，開所聞擴所見者多矣。②

序寫於至正辛丑秋九月望前二日（1361年 10月 12日）③。十五年後，陶宗儀的

另一位朋友，明初名儒、《元史》總裁官宋濂在《書史會要敘》中如此述陶宗儀

的編纂工作：

九成名宗儀，積學能文辭，嘗覽雜傳記一千餘家，多士林所未見者。

因仿曾慥《類説》，④作《説郛》若干卷。曾所編者，則略去之，君子謂其尤

精博云。⑤

易言之，這部書基本上是陶宗儀的一套讀書筆記，因此帶有他個性的烙印。

① 揚雄（前 53—公元 18）《法言》卷四：“天地萬物郭也，五經衆説郛也。”因此“説郛”之名表示它收

録了衆多作品，皆在五經範圍之内。見 Mote, T’ao Tsung-i and His Cho Keng Lu, pp. 79, 100; Paul

Pelliot,“Quelques remarques”,p. 163, n.1.

②《説郛》，上海：商務印書館，1927，《説郛序》（其二），第 1a頁；《説郛三種》，上海，上海古籍出版

社，1988，第 3卷，第 1頁；Mote, T’ao Tsung-i, p. 79. 我已將後出諸本所記之“一百卷”替换爲毛本中所

見的最初的“六十卷”；見徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，《東南文化》1994年第 6期，

112頁，116頁。

③ 日期僅見於毛本。參看徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，112頁，116頁。

④ 曾慥（1091—1155）《類説》是最早的類書之一。曾慥還編纂了道教類書《道樞》108卷。

⑤ 轉引自昌彼得，《説郛考》，10—11頁，以及徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，117

頁；部分英譯見Mote, T’ao Tsung-i, p. 100。404


惟一保存了《説郛》1361年原初形態的抄本是 60卷毛本①，現存浙江省臨

海市博物館。徐三見 1994年對其做了著録，并揭出其全目録②。它有三個顯

著特性：粗疏，簡短，特有的編排方式。毛扆 1710年云其“紕漏百出”“幾不可

讀”。後來的藏家王舟瑶（1858—1925）稱其“唯俗手所抄，誤字如麻，幾不可

讀”③。該抄本通篇遍佈通假字，常常以作者的方言爲之，如兒爲立，覺爲角，

治爲活，等等④。儘管有這些訛誤，毛本對於版本研究仍有重要的價值⑤。這

也是目前所知最短的完整本《説郛》，只有60卷、366目。

毛本也是《説郛》抄本中惟一具有連續性編次的。不像其他類書那樣按

主題分類，毛本按照各個作品題目的末尾一兩個字分類。題目尾字通常是

“筆記”、“録”或“傳”，這樣他的做法發展成一種天然的文體分類。卷 1—14所

收爲前人所編的各種叢書，卷 15—16爲“經”，17—18爲“史”，19—20爲“編”，

然後是“譜”“抄”“筆記”“紀聞”“談”“事”“話”“説”“志”“記”“録”⑥。《聖武親征

録》應該在最後一部分，但是像後來《説郛》所收的許多其他作品一樣，《聖武

親征録》不見於毛本。顯然元抄本不是《説郛》的最終版本。

β系统與γ系统

現今已知的所有《説郛》抄本，除毛本外，全部或至少部分内容所基之範

① 我使用這一簡稱因其藏於毛晉（1599—1659）的汲古閣。徐三見稱之爲汲古閣抄本，但是爲了

與賈敬顔取最早或最著名收藏者姓氏單字爲簡稱之法相符，我更願意稱之爲毛本。該抄本的出處，見

徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，112頁。毛晉在《南村輟耕録》跋中提及陶宗儀著作有

“《説郛》百卷，未能卒業”。見陶宗儀，《南村輟耕録》，北京：中華書局，1997，385頁；顯然他如此斷言是

基於廣爲提及的100卷與他所藏的60卷之間的矛盾。

② 徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，112—127頁。對此抄本的著録，以及部分書頁

照片，載周向潮、徐三見編，《歷史文化名城临海》“善本秘籍《説郛》”章，杭州：浙江人民出版社，2002，

245—248頁。

③ 徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，113頁、115頁。

④ 徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，116—117 頁。關於通假字，見 Wilkinson,

Chinese History, pp. 421-23, and Chinese History: A New Manual, pp. 45—46。

⑤ 吴晶，《〈説郛〉本〈洛陽伽藍記〉的版本價值》，《南京師範大學文學院學報》2009年第 1期，15—

17頁。吴晶注意到，儘管有一些明顯的訛誤如“萊”作“蒙”，但毛本：1）保存了他本誤、脱的正確文字；

2）有助於校正《洛阳伽蓝记》兩大系統的異文；3）較1927年商務印書館版（基於γ系統抄本）《説郛》有不

少文字更優。

⑥ 徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，113頁，以及118—127頁表格。 405


本，皆派生自一部 725目 100卷的《説郛》抄本，我稱之爲γ本（recension）。作爲

明代最爲通行的《説郛》抄本，γ系統諸抄本，都有相似的組織結構與内容，而

與 α系統的毛本很不相同。將 α系統的代表毛本與 γ系統的各種範本

（exemplar）進行比較就會發現，α系統的 60 卷内容全部塞入了γ系統的 30 卷

中。γ系統新加入了 70卷内容，使所收作品數由 366增至約 725①。在新加入

的内容裏，以題目末尾一兩個字分類的舊的編排方式基本上被無視了，甚至

從舊的《説郛》繼承而來的前 30卷也被打亂了順序。結果，γ系統的内容就混

雜着以主題、文體、題目尾字爲序的多種排序方式②。γ系統的這些範本仍然

有楊維楨序，但是删掉了日期，而且把序中的“六十卷”改成了“一百卷”以適

應新版本的篇幅。

γ本與α本相似的地方是頻繁出現簡體字和異體字。尤其是趙本、張本、

傅 3本，從譜系證據上來看相對較爲原初，保留了大量簡體字。雖然毛本中通

假字和直接的錯訛不常見，但簡體字方面，使用得非常普遍的有俻（偹）、称

（稱）、帰、难、禽（擒）、听，其他使用不一致的有报、边、抚、还、摊、无、、囙、

号、尽、惧、執、执、虽、与、捴。跟使用俗體字一樣突出的還有用字的不一致

性，即使在單獨一部作品如《聖武親征録》之中也是如此。

現存γ系統抄本中除了一種以外都是 16世紀或以後出現的。趙本爲弘治

庚申年，傅 3本（組成近全本《説郛》的三四種殘本之第三種，爲 20世紀學者傅

增湘首次著録）爲弘治十八年（1505）③。幾乎可以肯定比以上二者都早的，是

傅增湘《説郛》三四種抄本的第一種，含卷1—15。我簡稱之爲傅1本。這個抄

① 詳細比較了毛本與張宗祥100卷本所反映出的γ本《説郛》，我發現在毛本的366目或子目中，只

有 6個能在γ本卷 30之後找到。同樣，γ本前 30卷的 281目中，只有 36個（與《百川學海》重出者不計在

内）不是源出毛本。（書目總數的差異，取決於在具名的叢書中收録的選本是單獨計算還是只在叢書的

大題名下計算）其他γ 系統抄本皆與張宗祥本大體相同，包括張本（商務印書館編，《涵芬樓燼餘書録》，

上海：商務印書館，1951，第三卷，子，57b—63a）和臺本（《“國立中央圖書館”善本書目》第 2版，卷四，

1445—1484頁）。

② 在毛本中，前 14卷爲叢書——小部頭的叢書現在收録於大叢書之中。在γ系統中，原來在卷

34—35的儒家的“經”，被置於卷 1—2。後來，一位佚名編輯者重編 100卷《説郛》（ε本，其範例爲沈本、

史本），將所有的“經”——其中許多是晚近撰成的關於品味的著作（酒經、打馬圖經等等）而非儒家經

典——置於卷1。

③ 見本文附録對4種傅本的著録。根據我在上海圖書館的考察，傅1本無序。406


本寫在吴寬（1435—1504）叢書堂紙上①。饒宗頤推測這個抄本完成於成化八

年（1472）吴寬中進士之前數年②。鑑於所有的傅本都有同樣的編次（這使它

們能夠綴合成一部接近完整的《説郛》），這可將γ系統上溯至 15世紀中後期。

而且，趙本與傅 3本的《聖武親征録》截然不同，在它們與它們的祖本之間一定

經過了數次抄寫，這又一次將它們的祖本上溯至 15世紀。因此，從 60卷本擴

展爲γ系統至遲是在15世紀中期。

包括《聖武親征録》在内的一些作品，被抄寫進γ本是在元明鼎革早期，當

時陶宗儀仍然自命爲忠元之士。陶宗儀將每部作品加上了作者名（如果可

知）和作者的朝代。γ系統抄本中的幾部元代作品的作者是“皇元”。有兩部

作品的作者是“宋末國初”③。在《聖武親征録》的文本中，提及蒙古皇帝時總

是有空格提行，這可能是原作的特點，而陶宗儀抄録時小心地保留了下來。

從這些特徵判斷，這些作品應該是陶宗儀仍將自己看作元人時加入的。

卷 40到 60的所有元代作品也是如此，只有《困學齋雜録》的作者僅寫“元”④。

從卷 64以後的元代作品，都只寫“元”。然而，奇怪的是，卷 40之前的那些作

品，包括那些在毛本中已經收入《説郛》的，只寫“元”，雖然陶宗儀那時確實是

在元朝治下⑤。我的猜測是，卷 42到 57有“皇元”的作品是陶宗儀在特别關注

於强調自己元朝認同的時期加入的，非常可能是在他的妹妹與弟媳死亡的

1367年到明初的一兩年。這一時期他應該在收集《説郛》卷40—60的材料。

① 關於傅增湘（1872—1950）本《説郛》如何將殘抄本合併，最佳的材料是莫友芝著，傅增湘增補，

傅熹年編，《藏園訂補郘亭知見傳本書目》，北京：中華書局，1993，第二卷，10B/751—752頁。1927年版

的張宗祥跋中的著録（《説郛》[1927]，跋/1a；《説郛三種》，1358c）是不太準確的；更好的是張宗祥《説郛

校勘記》“説明”，載《説郛三種》，上海：上海古籍出版社，1988，1頁。

② 見 Jao Tsong-yi[Jao Tsung-i],“Un inédit du Chouo-fou: Le Manuscript de Chen Han de la période

kia-tsing (1522-1566)”,in Mélange de sinologie offerts à monsieur Paul Demiéville，Paris: Presses

universitaires de France, 1966, vol. 1, p. 93.

③ 首揭此者爲King P’ei-yuan，p. 4。“皇元”出現於以下作品（括號内表示 1927年版的卷數與頁

碼）：《春夢録》(42/18b)、《安南行記》(51/18b)、《聖武親征録》(55/1a)、《安雅堂酒令》(56/1a)、《鯨背吟集》

(57/1a)。“宋末國初”出現於《故蘇筆記》(57/20a)與《雪舟脞語》(57/20b)。

④ 見1927年版《説郛》，52/17a頁。

⑤ 見《佩楚軒客談》（7/22b；毛本，卷一一），《畫鑒》（13/1a；毛本，卷六）；《遂昌山樵雜録》（19/6；毛

本，卷五八），《浩然齋意抄》、《浩然齋視聽鈔》（20/1a 及 7a；卷二四），《山房隨筆》（27/12a；毛本，卷二

八）。有的作品的作者時代有矛盾。例如《錢塘遺事》（7/29a；毛本，卷三一）在毛本中繫爲宋，在 100卷

諸抄本中爲元。我只計入在毛本、100卷本中皆繫爲元的。 407


其他的内容，肯定是更晚才加入《説郛》的。所有的γ 系統抄本都收録了

幾部明朝作品：

1.《錢譜》，明佚名，述及永樂（1402—1424）事，昌彼得認爲成書於洪熙

（1424—1425）或宣德（1425—1435）初；卷84。

2.《古格論》，明曹昭，洪武二十一年（1388）；卷87。

3.《勸善録》，明仁孝皇后徐氏（1362—1407，1402年立爲皇后），永樂三年

（1404）；卷97。

4.《效顰集》，明趙弼，初稿成於永樂中期到宣德三年（1427），全書可能成

於正統元年（1436）；卷97。①

在 1、3、4作品的時代，陶宗儀肯定已經去世；因此由現存諸抄本所推證出

的γ本肯定不是陶宗儀編成的，而是有人續編。欲尋此人有一條綫索。在存

世γ系統最早的抄本傅 1本中，有關於編輯活動的題署：“南村真逸陶宗儀纂，

南齋龔鈇校正”。這一題署在張宗祥 1927年版本中得以重印。但我没有查到

這位龔鈇。他有可能是創造γ本的人嗎？這一問題仍有待進一步研究。

儘管如此，内部、外部證據皆表明，《説郛》從毛本的 366目擴展至明中期

的 725目一定不是一步完成的，而是分兩步。换言之，在α系統和γ系統之間還

有一個已經佚失的β系統。陶宗儀同時代人的記載證明他確實編成了 100卷

本《説郛》，不過這 100卷並不像 16世紀的 100卷一樣收録那麽多作品。楊維

楨稱讚《説郛》初編本收録作品“千餘家”，是他的過度誇飾，因爲初編本只有

366部作品。但在陶宗儀生前，他的《説郛》確實達到了 100卷，幾乎兩倍於初

編本。孫作1374年撰《陶先生小傳》，述陶宗儀著作如下：

① 渡邊幸三，《説郛考》，230頁；King P’ei-yuan, pp. 5—6；Jao Tsong-yi,“Un inédit du Chouo-fou”,

p. 94. 關於該作品撰成時間，見昌彼得，《説郛考》，366、370、386、388頁；以及徐三見，《汲古閣藏明抄六

十卷本〈説郛〉考述》，115頁。這些明初作品的存在使張宗祥很尷尬，因爲他曾宣稱 1927年商務印書

館 100卷本是基於録自陶宗儀稿本的洪武抄本。後來他注意到 1927年版確實含有永樂年間（1403—

1424）的作品，這只能是陶宗儀死後才被加入的。他的意見是：“或者原書早佚，明人隨意取他書補綴，

以成百卷之數乎？”張宗祥，《鐵如意館隨筆》，《中華文史論叢》1984年第 1輯，轉引自徐三見，《汲古閣

藏明抄六十卷本〈説郛〉考述》，115頁。對《説郛》本《效顰集》作詳細的文本分析，大概能根據帶有年代

的相關著作稿本來判定其成書時間，進而判定其編入《説郛》的最早時間。408


晚益閉門著書，世所共傳《説郛》一百卷，《輟耕録》三十卷……①

相似的，葉盛（1420—1474，1445年進士）《水東日記》提及陶宗儀：

近聞《説郛》百卷尚存其家，有九成塗改去取處，不知如何，其亦未成

之書歟？②

這些記載證實有一種 100卷本《説郛》是陶宗儀自己編成的，但今已不存，我稱

之爲β本。當然，如葉盛所説，《説郛》就是陶宗儀個人的讀書筆記，在他生前

並没有一個完整、固定的形態。葉盛的描述反映出編輯工作持續到陶宗儀死

前，也就是1401年以後不久。小的改動可能很頻繁地出現在β本中各處。

陶宗儀去世（1401年以後不久）之後，陶宗儀原編的《説郛》β本被壓縮成

不到 70卷，新的 30卷左右的内容被加了進來，這成爲了明中期的標準本——

725目、100卷的γ本。惟一提及這第二次重編的，是 15世紀作家都卬——其

生活年代僅可從其子、更爲知名的都穆（1459—1525）來推斷——都卬《三餘

贅筆》稱：“《説郛》本七十卷，後三十卷乃松江人取《百川學海》諸書足之。”③鑑

於《百川學海》的許多内容確實能在γ系統抄本中找到④，那麽肯定是某個“松

江人”（可能是陶宗儀家族中保存有葉盛所見抄本的人，也可能是傅 1本提到

的那位神秘的龔鈇）創造了《説郛》γ系統的第一個範本，γ系統《説郛》在 16世

紀便通行於世。

① 轉引自徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，115—116 頁；英譯參看 Mote, T’aoTsung-i, p. 31。

② 叶盛著，魏中平點校，《水東日記》，北京：中華書局，1980，71頁。轉引自昌彼得，《説郛考》，19

頁；饒宗頤，《香港大學馮平山圖書館藏善本書録》，香港：龍門書局，1970，160頁；參看法文譯本：Jao

Tsong-yi,“Un inédit de Chouo fou”, p. 89。

③ 轉引自饒宗頤，《香港大學馮平山圖書舘藏善本書録》，159—160頁。《四庫全書總目提要》沿用

其説，上海：商務印書舘，1933，123/2584；cf. Pelliot,“Quelques remarques”, p. 175；昌彼得，《説郛考》，

13頁。《百川學海》是 13世紀宋代叢書，收録作品皆爲全文，首次刊印於明代。按：今本《三餘贅筆》無

此記載。

④ 昌彼得，《説郛考》，15頁。 409


昌彼得對γ本《説郛》的徹底分析，證實都卬的説法是基本準確的①。γ系

統②有 725目。當然，昌彼得找到 72種與《百川學海》重複者，大部分在卷 67之

後③。這種分佈表明，γ本到卷 67之前與陶宗儀原編的β本大體相同，而β本所

收作品數大概在 572到 649之間④。陶宗儀所編與後人所編的分界綫在卷 60

和 70之間的某處，作爲印證的是明代作品出現於卷 84及之後，而作者題名表

現出忠元特徵的最後一部作品在卷57⑤。

以上探討的一個結論是，陶宗儀將《聖武親征録》抄入《説郛》β本，時在

1361年與 1374年之間。1361年是《説郛》α本産生的年代；1374年則是孫作撰

成陶宗儀小傳的時間，當時他已知道有 100 卷本《説郛》，即β本。後來“松江

人”重編增補《説郛》而創造γ本，則對《聖武親征録》没有影響。既然《説郛》産

生於元明交替前後而且收録了關於元興的作品，那麽從陶宗儀的讀書筆記來

推測其中反映出的他對王朝更替的觀點，便饒有興味。收入《説郛》的第二批

作品包含更多與邊疆、海外問題有關的作品（見附表）⑥。陶宗儀試圖理解作

爲非漢族王朝的元朝的遺産以處理它的覆亡嗎？抑或，藏書家賣出明朝人不

再感興趣的“蠻夷”題材書籍時，陶宗儀進行了大肆購買？可注意第二次編纂

① 我使用了昌彼得對 100卷本《説郛》中與《百川學海》重複内容的分析。但是我不同意他對數據

的解釋。我相信，他的錯誤來自於他誤認爲郁文博的編輯工作（見明清之際刻本中的郁文博序）與標

準的γ本有關。實際上，並不存在這樣一種有郁文博序的《説郛》。昌彼得的依據是臺本中有“都（郁）

文博校正”字樣，但我的文本分析證明，臺本毫無疑問是 20世紀製成的，其中反映出了當時的學術成

果，郁文博的署名很可能是那時加上去的。因此，根據臺北來判斷郁文博本的特徵是不可靠的。如果

不帶任何先入之見來讀郁文博序，顯然它不適於標準的γ本，而適於另一種《説郛》版本，估計郁文博並

非如昌彼得所指責的那樣不誠實。

② 昌彼得使用了臺北和張宗祥 1927年商務印書館本。我考察過的趙本、鈕本、張本内容也大體

相同。

③ 昌彼得《説郛考》，13—22頁，尤其是 15—16頁。卷 1—67的 580目中有 8種能在《百川學海》中

找到，卷 68—100的 145目中有 64種能在《百川學海》中找到。徐三見以張宗祥的研究爲基礎，也得出

了相似的結論，見《汲古閣藏明抄六十卷本〈説郛〉考述》，115頁。

④ 如果我們假設陶宗儀原編的所有作品都被移動到卷 1—67，那 580部作品中有 8部亦見於《百

川學海》，因而是後來插入的，所以我們得到 572之數。如果我們用 100卷本《説郛》的 725部作品，直

接減去明代的 4部作品和見於《百川學海》的 72部作品，就會得到 649之數。真正的數目大概在二者

之間，但更接近那個小的數。

⑤ 見前文明代作品列表。最後一部有忠元紀年的作品是在卷 57/20b（《雪舟脞語》，“宋末國

初”）。其後只有普通的“元”紀年的作品是《續積善録》（64/5a）、《景行録》（64/6b）。

⑥ 其中只有 3部作品，卷 86的《契丹國志》《大金國志》、卷 97的《遼東志略》可能是在將作品數由

600左右增至725的重編中加入的。410


相對而言收録了大量元代作品（《説郛》初編中很少有元代作品）①。這裏也可

能推測到的是，隨着王朝的衰亡（不論是即將發生還是已成事實），是否陶宗

儀試圖冒險保存它的不太知名的文獻。最後，可推測到的是，在第二次編纂

時收録《蒙韃備録》與《聖武親征録》，是否應部分歸因於陶宗儀認識到，儘管

他仍殘存忠心，元朝事實上已經覆亡且它的禁忌再也不會執行了。《蒙韃備

録》是 1221年宋人的作品，從宋人的視角描述蒙古，有時相當負面，通篇稱蒙

古人爲“韃靼”。而“韃靼”這一詞彙在元朝官方史書中通常都會被替换掉②。

另一方面，《聖武親征録》是實録的簡編本，只要元朝存在就只允許官方

讀者閲覽。這兩部作品在元代是不適合在大衆中流通的。儘管將佚名作者

尊稱爲“皇元”的作家，但在王朝更替之際收録這些作品有力地證明了《説郛》

是元亡以後的書。

拋開關於陶宗儀作爲讀者、藏書家、編纂者的推測，關於《説郛》中所收

《聖武親征録》抄本更重要的問題是，它是否爲全本，以及它的是否可靠。宋

濂説陶宗儀“曾所編者，則略去之”。《聖武親征録》也做過删略嗎？如果是，是

怎樣删略的？考慮到在成吉思汗的部分有一些幾乎完全不通的語句，看起來

陶宗儀完全複製了他所掌握的所有文本，縱然有不能理解之處（這是他忠於

元朝或對異國奇聞感興趣的標誌？）。窩闊台部分，文本更不清晰；考慮到對

窩闊台末年的記載是極其概略的，有可能是陶宗儀在這個節點做了删節。我

認爲更可能的是，陶宗儀做了全本複製，而窩闊台時期的删節是他所根據的

材料本身就有的。但陶宗儀大概確實做過删削，第二卷——名爲《親征録》，

記載了從貴由、蒙哥到忽必烈合罕的統治——整卷被删掉了，只是它的標題

出現在了《聖武親征録》的標題中。

如我下面將歸納并在抄本的章節中長篇討論的，《説郛》所收的《聖武親

征録》文本經歷了持續而層累的錯訛（corruption）。許多錯訛是隨機的，但是

① 考察涵芬樓本（商務印書館版）、臺北目録中卷 1—30和 30—67的作者時代可證。前者見《東

洋文化學院京都研究所漢籍目録》，310—321頁，後者見《“國立中央圖書館”善本書目》第 2版，卷四，

1445—1484頁。

② 關於“韃靼”、“達達”之使用的複雜問題，見蔡美彪，《元代文獻中的達達》，《遼金元史考索》，北

京：中華書局，2012，207—214頁。 411


也有許多是不斷地與《元史》文本進行呼應（harmonization）造成的。但有時，

對照《元史》《史集》，會發現即使《聖武親征録》的原型（archetype，指今可還原

的最古文本）也已有顯著錯訛。大多數錯訛看起來顯然是抄寫失誤，而有些

是爲與《元史》呼應，或試圖改進文本①。最早的毛本，如前所述，極其潦草，某

種程度上幾乎不可卒讀。如果《説郛》第二次編纂收録的作品也是這樣抄寫

的，那麽《聖武親征録》的許多錯訛就不應歸咎於明清的複製者，而是在《説

郛》版本傳承的最初就出現了。

因爲成書日期不確定，就還存在一種可能性，不能被證實，也無法否定。

《聖武親征録》抄入《説郛》是在 1361 年到 1374 年之間的某個時間。與此同

時，1369年到 1370年，明朝史官正以實録爲基礎編修《元史》②。從時間上，陶

宗儀作爲《元史》總裁官宋濂的好友，在將實録文本抄入《説郛》時並非不可能

注意到《元史》。這大概能説明一些與《元史》呼應的最早的例子，例如，《聖武

親征録》中插入的契丹、女真人的姓，或者《聖武親征録》所有抄本都与《元史》

有相同的錯訛，例如忽蘭·盞側（《元史》1.7，《聖武親征録》§15.3）應爲忽蘭·虎

惕③。在最早的諸抄本中可見到兩個用字不同的例子，相當可能是與《元史》

的呼應。Sa’ari（薩里~撒曆）、Küchülüg（屈出律~曲出律）這兩個例子，一種轉

寫用字不見於《聖武親征録》他處，但是與《元史》相符，另一種轉寫用字則在

《聖武親征録》中很普遍④。我傾向於認爲，陶宗儀抄寫《聖武親征録》入《説

郛》時親自將其底本與《元史》做了比較。這意味着他將《聖武親征録》抄入

① 按照我的《聖武親征録》校注，我認爲是較早的呼應。而大多數其他的注和校勘之處，以《實

録》之異文證之，皆爲錯訛。

②《元史》附録《進元史表》，4673—4674頁。《宋濂目録後記》，4677—4678頁；參看“Introduction”,

to Cleaves’s translation of The Secret History of the Mongols, Cambridge& London:Harvard University

Press, 1982, pp. xlv—l。

③ 當然，後一個例子不能排除有另一種可能，即實録本身就已經出現錯訛，然後被《元史》和《聖

武親征録》所沿襲。

④ Sa’ari 見於《聖武親征録》§§3.1、14.4 和 16.1，三者都能在《元史》1/3、1/6 和 1/7 中找到平行記

載。《元史》一律作“薩”，《聖武親征録》§§3.1 與 16.1 作“薩”，§14.4 則不然。Küchülüg 見於《聖武親征

録》§§33.2、36.2和 47，前二者在《元史》1/13和 1/14中有平行記載（參看《元史人名索引》，458頁）。《元

史》1/13和 1/14使用了“屈”；在《聖武親征録》最早期的抄本中，這只用於§33.2，他處則用“曲”。晚出

抄本爲與《元史》呼應，將多數或全部的“曲”改爲“屈”。第一種使用“屈”的情況，也可能是呼應的

結果。412


《説郛》只能是在1370年以後①。

γ系统與明中期書業

成化年間（1464—1487）開始，明代的書籍製造持續增長直到明末而延續

至清。這一漲勢在正德年間（1505—1521）積聚動力，到嘉靖年間（1521—

1566）書籍的刊印與製造所達到的數量級水平已高於明初②。雖然《説郛》到

明末才得以刊刻，但它以抄本的形式參與到了這次大發展之中。除了毛本和

殘缺不全的傅 1本，今存其他的《説郛》抄本都是弘治（1487—1505）及其後之

物。弘治年間開始，編輯者也開始嘗試用新方法改進《説郛》文本并重新編排

其結構，造成了抄本的混亂，延至今日。

“改進”《説郛》文本最常用的方法之一是找到其他範本，從中擷取“佳

處”。這些其他範本可能是收於其他《説郛》抄本之中者，但也可能是獨立於

所有《説郛》系統者。因此，《説郛》所收《聖武親征録》常常被拿來跟《元史》卷

1和卷 2的平行記載相對照，遇到相異之處則以《元史》改之。1505年前的某

時，一位佚名編輯者大規模這樣做，但仍是在標準的γ本脈絡之内。編輯者所

處理的文本非常接近《説郛》張本和孫本；張本很有可能就是他製作過程中的

一個稿本。從他所用的底本中，他繼承了幾個“因同字而脱字”之誤③，以及一

個奇怪的錯訛，將書中超過一半的“都”字（通常用於轉寫蒙古語的-du或-tu）訛

寫爲“相”。這將蒙古語Ba’atur（勇士）的標準音譯“拔都”變成了意義不通的

“拔相”。

但是，以一個類似於現存張本的抄本爲基礎，編輯者通過呼應《元史》，建

① 党寶海先生讀了本文初稿後提出：“上述改動也可能發生於明代中期抄寫《説郛》時所做的改

動。按照類書的編纂方法，編者通常不改動原書。陶宗儀應該能做到這一點。”一般來説党老師的意

見是對的，但是我估計像《百川學海》那種被發行的編纂文集跟像《说郛》這樣的個人使用的著作是不

一樣的。陶宗儀個人做的旁寫註釋很可能被以後抄寫的人抄入了正文。

② Lucille Chia,“Mashaben: Commercial Publishing in Jianyang from the Song to the Ming”, in TheSong-Yuan-Ming Transition in Chinese History, ed. Paul Jakov Smith and Richard von Glahn，

Cambridge: Harvard University Press, 2003, pp. 303—306.

③ 校勘學術語 parablepses，希臘文，相當於英文的 oversight，指視綫從一個字跳到下一個同樣的

字，因此遺漏了一大段文字。在複製《聖武親征録》這樣的比較難讀的文獻時這是很普遍的現象。（譯

者按：此即陳垣先生所舉之“因同字而脱字例”。陈垣，《校勘學釋例》，北京：中華书局，1959，25頁） 413


立起一個非常不同的體系。不僅删除了對元代皇帝頭銜的空格提行，編輯者

也注意到“乙亥”年（1215）的記載在文本中是缺漏了的。因此他找到《元史》

中那一年的記載，將這 121個字直接插入了§41之末。他没有注意到，或者可

能不在意，這些事件在《聖武親征録》的其他地方可以找到，只是版本不同①。

他也改動了轉寫，使它們與《元史》更像，將“王·可汗”改爲“汪罕·可汗”，“木

花里”改爲“木華黎”，“按攤”改爲“按弹”，“奕都·護”改爲“亦都·護”，等等。

編輯者通篇做了上百個這樣的小改動，有時甚至改錯了，但是許多時候通過

呼應《元史》使難讀的部分易懂了。編輯者還别具一格地將約半數的“真”改

爲“貞”；幸運的是這樣做不改變發音。

不管是出於使《聖武親征録》更可讀還是其他與編輯無關的原因，這樣改

動之後的《説郛》抄本——我稱之爲Hr（harmonized）範本——相當成功，産出

了大量的子抄本②。實際上現存《説郛》諸抄本中只有三種（趙本、鈕本、張本）

顯示出没有受到這個Hr範本的影響。1505年的傅 3本，顯示出這一編輯工作

的成果，并含有一些其他錯訛，所以這一編輯工作一定早於1505年③。

大概由於抄本的更大規模流傳，我們可以注意到將兩種《説郛》文本比對

而製作抄本的趨勢。一個編輯者複製了一種抄本，然後會將這個複製本與其

他抄本作比較，記下相異之處，如果看起來適當就將原來的替换。至少在《聖

武親征録》中，基本採取折中的方法，屢屢見到一段中採用了錯誤的寫法，下

一段又採用了正確的。有時候，比如汪本，是《聖武親征録》的一個Hr範本與

一個很相似於鈕本的本子進行校對，結果造就了一個得到大幅改進的文本，

① 這種不一致，在將《説郛》本《聖武親征録》呼應《元史》的嘗試中是很典型的。但洪煨蓮誤認爲

這些改動，比如 §41 之後插入的文字，是《聖武親征録》原本就有的。見 William Hung,“The

Transmission of the Book Known as‘The Secret History of the Mongols’”, Harvard Journal of AsiaticStudies, Vol. 14, No. 3/4 (Dec 1951), p. 480, n. 116.

② 明代抄本或多或少都是直接自此次編輯派生而出，包括傅 3本與涉本。20世紀傅—京本、臺本

也是複製自傅 3 本（或者臺北可能是複製自涉本）。我們可以稱這些本子爲傅—涉系（Fu-She

family）。《聖武親征録》潘本也是複製自兩個範本，其中之一就是由Hr範本派生而出的。

③ 在先前的文章中，我推測這可能是郁文博弘治九年（1496）重編工作的一部分。見Christopher

P. Atwood著，向正樹譯，《陶宗儀『説郛』と『聖武親征録』『蒙韃備録』のテキスト伝承》，《東洋史苑》77

（2011年 3月），127—150頁。部分是因爲當時我還没有考察臺北抄本，相信了昌彼得，將這一編輯工

作與臺北抄本聯繫起來。下文我將證明，臺北抄本並非郁文博系統，而其他抄本也没有這些編輯上的

改動。因此，没有證據能將這一編輯工作與郁文博聯繫起來。414


就差更多其他抄本以及非漢文平行文獻的利用了。許多後起的抄本也試圖

對早期γ本中特有的俗字進行改良。因此舊字形的存在可以作爲明中前期

《説郛》的一個標誌。

δ系统

其他的編輯者從整體上重新編排了《説郛》，或者增收作品，或者去除他

處可見的多餘作品。重新製作γ本的最著名者之一當屬郁文博（1418 年生，

1454年進士）。據郁文博自述，他於成化辛丑（1481）左右在陶宗儀生前所居

的上海地區得到了一部 100卷本（推測是γ系統的一個本子）。考察了文本之

後，他注意到其中多有訛缺和重出。他的抄本屢爲當地官員（“司牧部使者”）

借去傳抄，粗心的抄寫者造成了更多的訛誤。致仕之後，他想重新校正抄録，

同時也決定删除《説郛》中與新刊《百川學海》中重複的 63 部作品，剩餘的内

容，他重新編爲100卷。序繫於弘治九年三月（1496年3月）①。

假設郁文博所據的底本爲γ本，我們從他的序中得出的結論就會是，他編

纂了一部大概 662目、100卷的《説郛》。因此，與 725目的γ本不能等同。若是

没有發生一連串不幸的歷史性變故，學者們早就能明瞭這個結論了。現存的

γ系統諸抄本，與郁文博本找不到什麽關聯。這些抄本只有楊維楨序，無郁文

博序②，它們 725目、100卷的編次毫無郁文博序中所述的重編跡象，而現存可

繫年的 725目抄本，尤其是 1500年的趙本和 1505年的傅 3本，文本差别過大

（至少《聖武親征録》如此），不可能是源出一個共同的晚至 1496年才製成的祖

本。加上傅 1本與吴寬（1505年卒）叢書堂有關，無可置疑γ本早於郁文博稱他

製成了新本的 1496年。所有這些都確證，725目的《説郛》從一開始就與郁文

博無關。

但郁文博序被放在了明清更替之際的刻本中，所以很不幸有學者推測郁

文博本是明朝的通行本（textus receptus），學者也因此推測，欲還原明中期《説

① 郁文博序存於《説郛》刻本中，見《説郛一百二十》，1—3頁，載《説郛三種》，重印於昌彼得《説

郛考》，13—14頁。伯希和的法文釋義，載“Quelques remarques”, pp. 170—174.

② 我（直接或以照片複製品的形式）考察了趙本、張本、臺本的序與目録。基於 100卷明代抄本的

張宗祥1927年商務印書館版，亦無（第二個“説郛序”，1a—1b頁)。 415


郛》應即還原郁文博本。這一推測，伯希和、景培元等學者的研究中有所暗

示，昌彼得則有明言。昌彼得在臺灣只能利用 1927 年商務印書館版和臺北

“中央”圖書館藏的一個抄本（no. 000525628）。從目録看，這個抄本是一個標

準的γ本，只有楊維楨序①。卷 1 首頁有“上海後學都[郁之訛]文博校正”題

記。從這一條題記出發，昌彼得得出一個影響深遠的結論：通行的γ系統抄本

皆源出郁文博 1496年抄本②。不用説，這意味着郁文博序是很誤導人甚至是

不誠實的，因爲他聲稱他已删掉的作品卻仍皆在他的本子之中！

這條郁文博題記的真正出處，是我對臺北抄本的《聖武親征録》文本的詳

細研究之後才顯露出來的。我將在我的《聖武親征録》校注本中詳細論述，臺

北抄本並不是明代抄本，而是 1926年以後以假設的明代權威範本③爲基礎、結

合王國維 1926年校注本而造出來的贗品。對於《聖武親征録》而言，這個結論

是無可争辯的。既然有這個事實，而且没有郁文博序，那麽至少臺北抄本的

第一卷是插入了郁文博題記的現代複製品。鑑於臺北抄本《説郛》的編輯者

很天真地試圖“改進”《聖武親征録》文本（以明代的傅 3本爲基礎，至少《聖武

親征録》如此），在整個前半部分都沿用了王國維本的校正，所以他也基於

1920年代學界關於明代《説郛》的共識，毫不猶疑地加入了郁文博題記④。

因此，看來現存惟一可以確證的郁文博本是《説郛》明清刻本（下文將討

論），刻本編輯者收録了郁文博序，因而一定曾利用過郁文博抄本。不幸的

是，刻本没有收録《聖武親征録》。但收録了《蒙韃備録》，我對照γ系統其他抄

本（趙本、鈕本、張本）進行初步考察之後認爲，刻本所收文本是獨立於γ 系統

的，其中保留了在其他γ系統抄本中已經失落的許多古寫。這意味着，如同刻

①《“國立中央圖書館”善本書目》第 2版，卷四，1445—1484頁。我要感謝臺北“中央圖書館”特藏

文獻組主任兼漢學研究中心副主任俞小明，以及印第安那大學東亞圖書館劉雯玲的幫助，讓我能看到

序、目録的照片。

② 昌彼得，《説郛考》，14—15頁，圖一。饒宗頤從昌彼得説，見“Un inédit du Chouo-fou”,pp. 92—

93。

③《聖武親征録》的權威明代範本是傅 3本。但是鑒於臺北抄本有楊維楨序，而傅 1本沒有楊維

楨序，造僞者一定使用了四種傅本（Fu1-Fu4）之外的其他γ系統抄本作爲底本。

④ 臺本也删去了卷 97的兩部顯眼的明代作品（卷 84和 87的兩部明代作品得以留存）；見《“國立

中央圖書館”善本書目》第 2版，卷四，1482—1483頁，參照 1480頁。鑒於這個抄本是——即使其自稱

也是——明中期抄本，恐怕這一點也不能證明其古，而是爲使其看起來更早更真的又一粗劣手段。416


本的序一樣，這個文本可能來自一個獨立的郁文博本《説郛》。如此，則郁文

博 1481年借録、1496年完成的抄本，也應該是獨立於現存γ系統抄本的。鑑於

没有哪個《説郛》抄本能確定是源出郁文博本並且收録了《蒙韃備録》，我的這

個意見仍然只是推測，仍有待對更多的文本進行譜系性分析。

不過，我認爲有一個抄本可能是抄自郁文博本《説郛》，即滹南書舍本。

這個抄本是惟一存世的δ 系統範本，100卷（僅存 55卷），但是這 100卷内容只

與標準γ本的前 60卷左右相符（見表 2）。目前來看，這個抄本與γ本的編排順

序相同，僅卷 71—72有些例外。這是工作中的抄本（working copy），有大量的

校對標記（〇和丶），頁眉、行間有很多注①。傅增湘曾概論之，稱讚其善②。鑑

於其收録永樂及其後的作品《勸善録》《效顰集》（滹本卷 80=γ本卷 97）、《錢譜》

（滹本卷 70= γ本卷 84），所以它不可能早於 15世紀中期。不幸的是，序和卷 1

佚失了，無法證實我將它與郁文博本勘同的猜測。現存的 55卷似乎也不包括

《聖武親征録》和《蒙韃備録》，所以我現在還無法用譜系法確定其位置。但

是，整體的編排、抄寫年代都與郁文博序中所述很相似。况且，就我能確定的

δ本中的所有作品而言，没有發現與《百川學海》重出的，如果δ本就是郁文博

本的話這正是被期望的。這一鑑定的根本證據，只有通過比對滹本和刻本的

文本才能找到。刻本確曾利用郁文博文并用了他的序，其中至少應該有一些

作品文本上與滹本較爲一致，而與γ本不同或可能較之爲優。另一方面，如果

進行了全面的文本考察之後，發現刻本與滹本没有這種一致性，那我將收回

我現在嘗試性提出的鑑定意見。

ε系统

另一種重編的《説郛》現存兩個抄本，一個在香港，一個在北京。香港的

抄本爲沈瀚（1535年進士）抄本，簡稱沈本，69卷，1970年饒宗頤首爲介紹，現

① 原文與校文同時存在，開啓了有趣的可能性。現存抄本正在與其他抄本校對嗎？若是，那個

抄本能確認是哪個嗎？寫於書社紙張之上，可能也意味着《説郛》的又一個不爲人知的刻本在某種程

度上正被考慮。關於這個抄本的許多問題仍有待研究。

② 莫友芝，《藏園訂補郘亭知見傳本書目》，752頁（所列第二種《説郛》）。 417


藏香港大學馮平山圖書館①。北京的抄本爲國家圖書館編號 no. A01507，簡稱

史本②。這兩個抄本中的《聖武親征録》極其接近，這兩個抄本組成了《説郛》

的ε系統。

ε系統的兩個抄本都没有目録和卷數。没有目録，因而不便於學者全面著

録其内容；迄今尚無人這樣做。饒宗頤著録過沈本部分卷帙的内容，而我也

介紹過史本殘帙的編排。二者都顯示出，其編排完全不同於標準的γ本（見表

3 和表4）③。僅通過比較大致内容所不能看到的一個突出的特點是，這兩個抄

本的頁面佈局是一致的，每行每字都在相同位置——如果其中一個抄本（通

常是抄寫較爲粗疏的沈本）遺漏了一個字，在行末就會重複一個字以保持頁

面佈局一致。從饒宗頤所述的部分内容諸如《諸子隨識》《諸傳摘玄》的子目

編排來看，沈本的編排相對於標準的γ本距毛本更遠④。ε本編輯者顯然開始傾

向於採取以内容爲基準的主題式編排⑤。而且，69卷之數與都卬所説的 70卷

很接近，而實際收書總數更接近γ本而不是毛本。

沈本、史本都含《聖武親征録》⑥。ε系統的這兩個抄本，文本也非常有特

色。相對於其他的《聖武親征録》，它有三個主要特點：1）它的底本是從Hr範

本的一個極近的後嗣中派生出的本子，這個大量呼應《元史》的γ系統範本也

被明代的傅 3本、汪本、涉本所用。2）這個 Hr範本，與一個今已佚失、獨立於

《説郛》流傳的《聖武親征録》文本校對過⑦。3）最後，它又跟《元史》對校過一

次，出現了其他抄本没有的、來自《元史》的 23條或大或小的插補，以及許多小

① 最早的著録見 Jao Tsong-yi 1966；饒宗頤，1970，更長的著録見 1982年的文章《〈説郛〉新考——

明嘉靖吴江沈瀚鈔本〈説郛〉記略》，《饒宗頤史學論著選》，上海：上海古籍出版社，1993重印。傅增湘

增訂莫友芝書時亦曾述之，見《藏園訂補郘亭知見傳本書目》，752頁所列最後一種《説郛》。

② 北京圖書館編，《北京圖書館古籍善本書目·子部》，1695頁。

③ 兩個抄本的編排有一些不一致。沈本中，《聖武親征録》之後即《北轅録》；而史本中，《北轅録》

在第 10册，而《聖武親征録》在第 11册。但鑑於史本殘缺太甚，沈本基本完整，我推測，史本的編次是

將零散的殘本重新裝訂的結果。

④ 見饒宗頤，1993，661頁。

⑤ 毛本中，陶宗儀將前 14卷收叢書。在後來的 69—70卷抄本中，陶宗儀將原來在卷 34—35的儒

家的“經”置於卷1—2。而ε系統將那些晚近撰寫的非儒家經典的“經”（酒經、打馬圖經等等）放入卷1。

⑥ 二抄本内容見表格。不幸的是，史本殘缺太甚，只有《説郛》原本的四分之一或五分之一，其中

没有什麽内容與饒宗頤對沈本的粗略著録重合。但是二者都與γ本迥然不同。

⑦ 這個原初的範本類似於另一個非《説郛》抄本，這個非《説郛》抄本也被陸本（東京静嘉堂文庫）

與鄭本（中國國家圖書館）《聖武親征録》的共同祖本使用過，似乎是《皇元聖武開天記》明代抄本。418


的改動。結果ε系統文本成爲最有特色的《聖武親征録》。偶爾出現的來自非

《説郛》本的異寫極爲有價值，但是常常被大量的呼應和獨有的異寫所掩蓋，

尤其是沈本在轉寫蒙古文方面選擇了幾個相差更遠的字，如“札”作“杞”，

“滅”“蔑”作“茷”。

ε本的最初編輯者和編成時間不詳。沈本成於嘉靖年間（1512—1566），但

從該系統的共同祖本到沈本之間至少還複製過一次。不論如何，ε 系統的文

本不是特别原初。如我所提及的，其中的《聖武親征録》顯然依據了Hr範本的

一個抄本，後者是 15世紀末 16世紀初γ系統的一個分支。同時，我對《蒙韃備

録》的初步校勘表明，相較γ 系統其他抄本，沈本與史本有最大量的共同的新

改動（共有衍徵，synapomorphies）（與張本大體接近，與趙本、鈕本較遠）。因

此，文本證據很明確：ε本的編排，不像饒宗頤所認爲的那樣，能證明自身的原

初性；而是取一個標準的γ本並重新編排得看起來接近陶宗儀原本的想法。

估計編輯者面對的讀者群已諳熟葉盛和都卬對文本原初形式的懷疑，可能也

諳熟郁文博的重編本。這幾位著名目録學家的論斷，爲書籍市場上推出 70卷

本《説郛》灌注了動力。對那些質疑《説郛》文本的人來説，ε本看起來要比其他

“增加蕪穢，妄置次第”①的《説郛》優異很多。事實上，是沈本的編輯者造成了

書目學上的錯亂。

ζ系统

鑑於《説郛》越來越受歡迎，有人想把它刻版就只是時間問題了②。1621

① 見沈本的藏家陸樵的評論（己丑，可能是 1589年），轉引自 Jao,“Un inédit de Chouo fou”, p. 91；

饒宗頤 1993，657頁。陸樵的評論是根據了都卬對 100卷本的疑問嗎？抑或陸樵也注意到了郁文博本

及其序？

② 最便於查閲的刻本載於《説郛三種》。《説郛三種》影印的是張宗祥 1927年版 100卷《説郛》、明

清之際 120卷刻本《説郛》及《續説郛》。但很重要的是，必須注意，《説郛三種》中所印的並不是任何一

部現存刻本，而是一個混合本，將明末刻本與清初序混合在了一起。影印本的實際内容與編排，與京

都東方文化研究所藏本——崇禎年間的初印本一致。因此它收録了後來的刻本所刪掉的内亞作品。

但是影印本也收録了順治三年（1646）李際期序、王應昌序，這只能是刪除了内亞作品之後才加進去

的。同樣值得注意的是，《東洋文化學院京都研究所漢籍目録》324ff頁著録京都刻本也有陶珽之名、

李際期序、順治三年款，實際上這些皆不見於該本，該刻本毫無疑問是崇禎本，此點渡邊幸三早已

斷定。 419


年前的某個時間，杭州的宛委山堂製作了首個《説郛》刻本①。刻者既然使用

了郁文博序，就定然利用了郁文博系統的一個稀見抄本（如我推測的，δ系統）

作爲刻本的主體，但是這個抄本大概不全，不包括讀者所期待的全部 725目。

所以，像其他《説郛》新版本一樣，刻者肯定用了多種抄本。結果最終版就是ζ

本，幾種抄本具體如何合併仍不清楚。

《説郛》諸抄本錯訛已經很多，所以他們儘可能從其他類書中找出諸作品

的刻本來使用。不斷地增收新作品，卷數增至 120卷，附加 47卷的“續”。楊

維楨序中的“一”字改爲“二”，不論是因爲蓄意還是出於錯訛，陶宗儀的“一千

家”變成了“二千家”！爲了完成這個杜撰的數目，刻者遍尋類書吸收新作品，

將長篇作品割裂，將同一作品加上不同的名字，甚至加了一些他們認爲陶宗

儀可能收録的作品並注以“闕”。刻本以主題編排，東、北行紀在卷55和56，邊

遠地區、東南亞紀聞在卷 62。因爲增至 2000部作品，所以新收了許多陶宗儀

未收的有關域外的有趣作品（見表1）②。

然而刻本付印之前，天啓元年（1621）的大火對杭州造成巨大破壞③。刻

版的全部或大部幸存了下來，但是宛委山堂已無財力將之付印，便將刻版售

於杭州的其他印社，版片稍作改動後分散付印而成爲六部其他類書。明末崇

禎年間（1628—1644），宛委山堂終於重新購回刻版，將《説郛》120卷及明人續

46卷印刷了兩次。收作品總數約爲 1360④。有郁文博序、楊維楨序，以及出版

者的《讀説郛》題記。γ本中的一小部分作品不見於任何刻本，其中包括《聖武

① 以下我完全依賴昌彼得精彩的探索性著作；昌彼得，《説郛考》，25—35頁。

② 例如，劉郁《西使記》記伊利汗國，石茂良《避戎夜話》記女真金，胡嶠《陷虜記》記契丹遼，宋延

德《高昌行紀》記回鶻，方鳳《夷俗考》、徐兢《使高麗録》記高麗。

③ 昌彼得相信有記載提及了 1621年以前的印本，但是這種印本未能留傳至今（見昌彼得《説郛

考》，27—28頁)。

④ 1943年北平中法漢學研究所（Centre Franco-Chinois d’Études Sinologiques）購入一部刻本，景

培元（King P’ei-yuan, Études comparative, pp. 6-9）曾做過探討，昌彼得將其定爲迄今所知最早的刻

本。全目録（基於 cat. no. 4104-87-3560）刊佈於《静嘉堂文庫漢籍分類目録》，966—990頁。東洋文化

學院京都研究所藏明代刻本，由渡邊幸三 1938年的《説郛考》刊佈，他認爲此本略有增補，至晚 1643年

刊印。其内容已發表於《東洋文化學院京都研究所漢籍目録》，324—347頁。初印本爲 120卷、1360目

（其中 124目註爲闕）及續 44卷（含 544目，其中 6目闕），第二次印本有 1364目（其中 124目闕）及續 46

卷（含542目，其中8目闕）。見昌彼得《説郛考》，30—31頁。420


親征録》①，不知是因爲所有抄本都有缺失還是這部分版片没有找回來。

明亡清興的動亂，使最早的刻本没能廣爲流通。而且，在滿清統治下，早

期《説郛》所收的那些反映宋人對契丹、女真、蒙古政權態度的作品很成問

題。一個謹慎的出版商不會冒險出版含有詛咒滿洲朝廷爲“虜廷”——以及

含更貶義稱謂——的作品。1646年宛委山堂重印《説郛》廣爲流傳，其中不含

《聖武親征録》《蒙韃備録》。

在崇禎年間的初印本中含有的 14部關於契丹遼、女真金、蒙元王朝的作

品，在清初的印本中只剩下 3部，或者最多 4部②。編輯者爲李際期（1640年進

士），有的印本中加入了李際期序和王應昌序，而其他印本中仍然使用了此前

刻本的序及卷首③。李際期的節本成爲清朝官修《四庫全書》（1772—1794）本

《説郛》的底本。直到 1927年，它一直是《説郛》標準本，儘管它跟陶宗儀原書

已經差别很大④。

《説郛》刻本定義了人們心目中的《説郛》，使進一步的抄本製作遲緩了下

來。不過，仍偶有抄本出現。1778年，鄭傑對照一個《聖武親征録》抄本，抄録

了他的一位朋友從蘇州閶門附近一位收藏家處所得的一部《聖武親征録》。

没有更多信息來確定這是《説郛》的哪個系統，但從鄭傑所録的異文來判斷，

這個《聖武親征録》文本不屬於任何一種《説郛》本⑤。這個有趣的版本證明，

① 在上述關於中國邊疆、域外的那些作品中，見於 100卷抄本而不見於早期刻本的只有《聖武親

征録》《使遼録》《青塘録》。昌彼得指出 100卷本的 725部作品中有 206部不見於刻本（30頁）。王舟瑶

1917年的跋云，以 1361年毛本與刻本相比較，見於刻本而不見於毛本的作品有 860餘種，而見於毛本

不見於刻本的作品有 100餘種。（轉引自徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，113頁）。若如

此，域外邊疆題材在明刻本的保留，佔了一個相當高的比率。不過，因爲當時缺少一個方便的索引

——今日可用《説郛三種》第10册的索引，他們的統計中很可能仍然有遺漏的作品。

② 此據我對東洋文庫的四個本子（cat. nos. V-5-A-11, 12, 13, 14）以及静嘉堂文庫的清印本（cat.

no. 8505-163-305-2）的調查。《説郛》刻本中的這種典型的刪節内容，見《静嘉堂文庫漢籍分類目録》，

990—1014 頁，據 cat no. 8505-163-305-2。參看 King P’ei-yuan, Études comparatives, p. 15。注意全版

從未被重雕；差别僅僅在於是否使用某個版片，在目録中挖去冒犯性的題目並粘上修正過的題目。

③ 東洋文庫 no. V-5-A-11 卷首一仍其舊，而東洋文庫 no. V-5-A-12 與靜嘉堂文庫 no.

8505-163-305-2 使用了有李際期、王應昌序的新版本。新版本卷首也插入了“姚安陶珽重輯”字樣。

昌彼得認爲，關於陶珽（雲南姚安人，1610年進士）生平的史料中無任何有關他重編《説郛》的記載，因

此未必確有其事（昌彼得《説郛考》，22—25頁）。

④ 1988年的《説郛三種》3—10册影印的是一部混合本，包括李際期印行時加入的序及卷首，以及

崇禎年間第二次印刷的《説郛》及《續説郛》。

⑤ 不過，它與南京圖書館藏汪本《聖武親征録》很相似。 421


《説郛》的版本流傳仍然有許多不爲人知之處，而且仍有許多殘抄本有待發掘

和鑑定。

1927年商務印書館版

前已提及，1927年商務印書館版《説郛》是一個里程碑，它使曾經居於主

導地位但在明末被ζ本取代的γ本重歸學界視野。張宗祥編輯的商務印書館版

《説郛》在内容上與張本基本相同①。

張宗祥實質上没有對他如何編輯該書留下多少信息，只在書末的跋中有

一個簡要的抄本列表。跋繫於壬戌（1922年），列出了其所用的六種 100卷本

《説郛》，皆非全本②。第一種是京師圖書館藏本，至卷 32；第二種是由三種抄

本匹配而成的傅本，今藏上海圖書館；第三種是涵芬樓藏本，我採用賈敬顔的

簡稱張本（因張元濟首爲著録，故名）；最後一種是孫本，按張宗祥的説法，他

僅以此本補他本之缺卷。易言之，張宗祥版本《説郛》中的每一部作品，皆基

於一個單獨抄本。據我所知，這幾種抄本，後三種今存，第一種已不存③。

賈敬顔已經注意到，關於張宗祥本《聖武親征録》有些疑問。首先他指

出，在張列出的所有抄本中，含《聖武親征録》的是弘治十八年的傅 3本，賈敬

顔繼云：

然此重排《説郛》本《聖武親征録》，既不與王國維所校而伯希和所載

之傅本相同，亦不復類似張本，文字優劣處，乃在二者間，究竟何所出？④

譜系研究解決了賈敬顔的疑問，1927年本並非基於一個抄本，而是合併

了張本及賈敬顔所稱之京本（賈敬顔以其寫於“京師圖書館鈔”書紙上，故名

① 這個版本流傳很廣，有1927年原版、1988年《説郛三種》影印版。

② 見《説郛》，1927 年版，《跋》，1a—b頁；《説郛三種》，1358c—d頁。

③ 這個抄本的内容可以還原嗎？某種程度上我認爲可以。下文我將提及，浙江圖書館有一部張

宗祥編輯《説郛》時曾用過的抄本。在這個抄本里，底本（偶有編輯者的校正）書以墨筆，根據他本而做

的校對書以朱筆。與張宗祥所用的其他現存抄本（傅 1-4本，張本，孫本）比較，與其他抄本明顯不同的

文本可以獨立出來。這樣的文本可推定是出自張宗祥所用的第一種抄本。

④ 賈敬顔，《聖武親征録校本》I，《綴言》，4a頁。422


之）①兩種抄本。後一種抄本是傅 3本的一個相當粗劣的現代複製本，偶爾以

1894年付印的《聖武親征録》最早的學術版本做過校勘。譜系研究的這一成

果，爲張宗祥的稿本所證實。今藏浙江圖書館的稿本（no. 7437）②保留了張宗

祥的底本，墨筆所書，很大程度上與傅 3 本相符，但有一些小的校正；朱筆所

書，是張宗祥加上的來自張本的異文。這個證據表明，1927年版本中並非如

人們按張宗祥跋可能推想的那樣每一部作品來自一種抄本。相反，它們通常

是合併了兩種或更多γ系統抄本的混合文本。這種混合性特徵，至少在《聖武

親征録》這一案例中是不太清晰的，因爲張宗祥所用的抄本——張本、傅 3本

相互之間就已經很接近了③。在《説郛》的其他作品方面，如果所用抄本互相

之間關係更疏遠，那麽1927年版的混合性特徵大概會更明顯。

結論

這裏的結論，只是對《説郛》所收作品文本研究的一個開始。這些結論，

基於對《聖武親征録》的詳細分析以及對《蒙韃備録》與楊維楨序的初步分

析。因此，仍然有許多問題未能解決，有些答案也只是暫時性的。包括δ本與

郁文博本的勘同，以及用以創造ζ本的具體是哪個抄本。

而且，無法保證《説郛》的一種特定抄本中與另一種抄本中的所有作品都

有相同的譜系關係。我對《蒙韃備録》的初步研究表明，它的譜系確與《聖武

親征録》相符。不過，《説郛》抄本中的一個作品被複製時不只是基於一個範

本而是基於兩個或更多的範本，這種現象並非罕見。我所考察的諸《説郛》本

《聖武親征録》中，傅 3本、汪本、史本、沈本都可證曾將底本與另一抄本對校。

换言之，製作抄本的學者曾利用不止一部《説郛》本，總是有可能他們對《説

郛》中的一個作品採用一個抄本，另一個作品採用别的抄本。而且，明代《説

① 傅—京本現藏國家圖書館，賈敬顔曾利用之，《聖武親征録》在卷 55；不能與張自己所用的本子

混淆，後者藏於京師圖書館，但没有卷32之後的内容。

② 全目録載浙江圖書館古籍部編，《浙江圖書館古籍善本書目》，杭州：浙江教育出版社，2002，

670—680頁。

③ 我在《聖武親征録》校勘本中將闡述，這兩個抄本非常接近，因爲編輯者在明弘治年間製作傅 3

本時，使用了張本或與之非常類似的抄本做過校對。 423


郛》抄本不太可能總是完整的。16世紀欲製作《説郛》全本者，可能與 20世紀

的傅增湘、張宗祥採取相同的權宜之計：用一些或長或短的殘本拼湊成 100

卷。對於這樣一部由幾種殘本合成的《説郛》，版本分析所得的譜系取決於選

擇其中哪部作品做分析。

因此，只有將《説郛》所收的約 725目每一個都做詳細的分析，最終才能看

到這部類書的發展史的全景。對每部作品的深入研究，一個益處是可以鑑定

中國各圖書館今藏的大量《説郛》殘本，有的僅存 5卷或 10卷甚至更少。從中

發現價值極高的抄本的範本殘卷，並非不可能。無論如何，鑑定必須逐個進

行，分析保存得更好、更爲知名的範本中派生出的作品，理解其譜系，是研究

的基礎。這一巨大的工作，顯然超出了任何一個學者的個人能力，只有在一

個目標下團隊協作，集合與陶宗儀所選所有題目有關的各方面學者方可。希

望本文的初步分析能對欲從事此項工作的學者有所幫助。

附録：《説郛》諸抄本

今存之抄本

1. 毛本：現藏臨海市博物館①。α系統。存 60卷，20册。不含《聖武親征

録》。抄本全目録已發表②。有楊維楨序；内容以題目末字排序，與 1927年商

務印書館版完全不同。無行格，無魚尾，版心無題記。每半頁 9 行，每行 17

字。27.3×17.5釐米。墨書校正可能是最初編輯者所爲，朱書校正爲後世藏家

所爲。

流傳史：通常認爲是明代抄本。字形上的不統一——尤其是陶宗儀特意

用來代替更常用的“卷”字的罕見字“ ”——大概意味着它距離陶宗儀 1361

年原本至少經過了數次抄寫。可知的最早藏家是毛扆，朱書校正，卷 20題記

及“虞山毛扆手校”印。後爲馬玉堂（字笏斋，號秋藥，1821年副貢，1845年賜

同進士出身）所得。同治年間歸王咏霓（1839—1916，字子裳，號六潭，1880年

進士）得之，存放於台州市黄嚴區的九峰書院（後爲黄嚴九峰圖書館）。其同

① 非常感謝徐三見先生在我2013年1月8日參訪臨海時允許我校對並拍攝此抄本的部分書頁。

② 徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，112—127頁。424


鄉王舟瑶（1855—1925，字玫伯，號默庵）以朱書做了進一步校正，在序後加了

一個跋①。

2. 趙本：現藏中國國家圖書館（no. 3907）。γ系統。存 61卷，50册。卷 55

爲《聖武親征録》②。有楊維楨序；内容與 1927 年商務印書館版僅偶有小差

異。藍格，雙魚尾，黑口，版心無數字、無題記。

流傳史：在卷 24 末，賈敬顔找到了“弘治庚申依本録”題記，因此抄録於

1500 年。有許多學者印章、題記：吴郡趙氏、阮元（1764—1849）、翁斌孫

（1860—1922）等人③。賈敬顔推測吴郡趙氏可能是趙宦光（萬曆年間人，

1572—1620）或是他的兒子趙均（崇禎年間人，1627—1644)。國家圖書館得之

於翁斌孫。

3. 鈕本：現藏中國國家圖書館（no. 2408）。γ系統。存 97卷，70册；《聖武

親征録》爲卷 55④。全目録已發表⑤。有楊維楨序；内容與 1927年商務印書館

版僅偶有小差異。藍格，無魚尾，白口；版心無數字，但有“世學樓”題記。世

學樓屬於明代藏書家紹興鈕緯（字仲文，號石溪，1518—1579，嘉靖二十年

[1541]進士），此抄本爲其所抄録⑥。

流傳史：賈敬顔記載，此書經常熟何棹、陳揆（字子准；1780—1825）收

藏。清代目録學著作《郘亭知見傳本書目》也提及此抄本爲陳揆所有。賈敬

顔又記，陳揆藏書多屬常熟錢謙益（1582—1664）舊藏，錢氏的私人藏書目録

① 見徐三見，《汲古閣藏明抄六十卷本〈説郛〉考述》，112—115頁；以及《善本秘籍〈説郛〉》，周向

潮、徐三見編，《歷史文化名城临海》，245—248頁。後者有印章、跋的照片。

② 北京圖書館善本部編，《北京圖書館善本書目》，北京：中華書局，1959，卷五，37a頁；北京圖書

館編，《北京圖書館古籍善本書目·子部》，北京：書目文獻出版社，1987—1988，1694頁；中國古籍善本

書目編輯委員會編，《中國古籍善本書目·叢部》，上海：上海古籍出版社，1989，1.38a頁（1990年重印本

中爲75頁）；翁連溪編，《中國古籍善本總目》，北京：綫裝書局，2005，卷六，集部（下），叢部，1927頁。

③ 賈敬顔列出了其他擁有者包括：睿鍾、徐鐵彜、趙元修、衛去疾，以及私人藏書館常熟捨莊、楊

氏善慶堂、周鑒齊、削漢劍、魏慰斗主人。

④《北京圖書館善本書目》，卷五，36b頁；《北京圖書館古籍善本書目·子部》，1694頁。

⑤《中國古籍善本書目·叢部》，1.10a—24b頁（1990年重印本中爲 19—48頁），以及《中國古籍善

本總目》，卷六，集部（下），叢部，1919—1923頁。

⑥ 感謝馬曉林幫助我找到鈕緯的生活年代和籍貫。 425


提及《説郛》100卷，賈敬顔以爲當即是書①。

4. 孫本：現藏浙江省瑞安市玉海樓博物館。γ系統。存 52卷，18册；卷 55

爲《聖武親征録》②。藍格，版心上部有暗淡的單魚尾；版心無數字，無題記。

正文墨書。

流傳史：明代抄本，藏於孫詒讓（字仲容；1848—1908）舊宅，爲張宗祥

1927年商務印書館版所用的第四種抄本。

5. 張本：現藏中國國家圖書館（no. 7557）。γ系統。存 91卷，29册；《聖武

親征録》爲卷 55③。全目録已發表④。楊維楨序；内容與 1927 年商務印書館版

大體相同⑤。藍格，單魚尾，白口；版心無數字、無題記。

流傳史：被列爲明代抄本。張宗祥繫之於萬曆時期（1572—1620），但我

認爲可能更早，早於傅 3本。賈敬顔記卷册只有校對者張元濟⑥印、涵芬樓印，

涵芬樓是以張元濟爲主編的商務印書館的藏書樓⑦。這是 1927年商務印書館

① 賈敬顔，《聖武親征録校本》（未刊稿本，乙未）I，《綴言》，3b頁；莫友芝著，傅增湘增補，傅熹年

編，《藏園訂補郘亭知見傳本書目》，第 2 卷，10B/751 頁；錢曾，《述古堂藏書目》，臺北：藝文印書館，

1965重印，2/17b頁。不過，如饒宗頤所述，國家圖書館藏鈕本爲 70册，而《述古堂藏書目》所記爲 32

册。如果二者爲一，那麽它應該重新裝訂過。見 Jao Tsong-yi,“Un inédit du Chouo-fou”, p. 93；饒宗

頤，《〈説郛〉新考——明嘉靖吴江沈瀚鈔本〈説郛〉記略》，《饒宗頤史學論著選》，659頁。

②《中國古籍善本書目·叢部》，1.38b頁（1990年重印本 76頁）；《中國古籍善本總目》，卷六，集部

（下），叢部，1927頁；上海古籍出版社“説明”，載張宗祥《説郛校勘記》，載《説郛三種》，上海：上海古籍

出版社，1988，1頁。《説郛》1927年版的《跋》，1a—b（《説郛三種》，1358c—d頁）不太準確，信息也不全

面。亦見“中國古籍善本書目聯合導航系統”在綫書目，説郛一百卷［明陶宗儀編明抄本］存五十二

卷［九至十、十六至十九、二十三至三十、三十三至三十七、三十九至四十四、五十、五十二至五十三、五

十五至六十一、六十三至六十六、七十至七十二、七十九至八十二、八十八至九十、九十二至九十四］

http://202.96.31.45/libAction.do?method=goToBaseDetailByNewgid&newgid=6193&class=kind

(accessed June 30, 2010)。

③《北京圖書館善本書目》，卷五，36b—37a頁；《北京圖書館古籍善本書目·子部》，1694頁；《中國

古籍善本書目·叢部》，1.38a 頁（1990 年重印版 75 頁)；《中國古籍善本總目》，卷六，集部（下），叢部，

1927頁。

④ 商務印書館編，張元濟序，《涵芬樓燼餘書録》，上海：商務印書館，1951，卷三，子，57b—63a

頁。含全目録。

⑤ 抄本卷首只有卷1—8的目録，無作者名，無作者朝代。

⑥ 關於張元濟，見 Manying Ip, The Life and Times of Zhang Yuanji, 1867—1959 ，Beijing:

Commercial Press, 1985。

⑦ 賈敬顔，《聖武親征録校本》（未刊稿本，乙未）I，3b頁。426


版所用的第三種抄本，被用來校對以傅3本爲底本的《聖武親征録》①。

6. 傅本：傅增湘（1872—1950）以三或四種殘抄本②匹配成一部近乎完整

的《説郛》。現藏上海圖書館（nos. 786660—786719）。皆屬γ系统③；《聖武親征

録》在卷 55。傅本全目録已發表④。傅 1本（卷 1—25）爲黑格“叢書堂”紙，每

半頁 10行；傅 2本（卷 26—30及 96—100）爲藍格“弘農楊氏”紙，每半頁 11行；

傅3本（卷31—67）與傅4本（卷68—70）爲“説郛”紙，每半頁13行。

流傳史：傅 1本寫於吴寬（1435—1504）叢書堂紙上。傅 2本年代不清楚，

因我還没確定“弘農楊氏”代表哪一時代的書院⑤。傅 3本卷 62有一行注：“弘

治十八年（1505）三月。”傅 1 本卷首有傅增湘“雙鑑樓”印。全書多見王體仁

（字綬珊，1873—1938）印⑥、上海圖書館印。這是張宗祥 1927年商務印書館版

所用的第二種抄本。傅3本是其中《聖武親征録》的底本。

7. 汪本：現藏浙江圖書館（no. 7434）。γ系統。存 41卷，26册；《聖武親征

① 此抄本與張宗祥跋中提到的“涵芬樓藏本”相同。張宗祥跋云：“一爲涵芬樓藏本，似係萬曆抄

者，未缺各卷，每數卷前有目録，今之目録即自此本寫定者。”見《説郛》（1927），跋，1a頁；《説郛三種》，

1358c頁。從這個著録看涵芬樓本是完整的，不缺卷。然而，張宗祥的這個著録是誤導人的。一個更

准確的著録載於《説郛三種》，言涵芬樓本是明抄本，殘 91卷，與張本的卷數相同。見上海古籍出版社

“説明”，載張宗祥《説郛校勘記》，載《説郛三種》，1頁。

② 對於是否存在一個獨立的傅 4本，目録學者有不同的觀點。傅 1、傅 2、傅 3的紙張與書法都明

顯不同。但傅增湘又分出傅 3（卷 31—67）、傅 4（卷 68—70）二本，儘管二者非常相似，皆爲黑格紙，每

半頁 13行，版心印有“説郛”二字。見莫友芝著，傅增湘增補，傅熹年編，《藏園订補郘亭知見傳本書

目》，第二卷，10B/751-52 頁。而張宗祥並不分出傅 3 和傅 4，將傅增湘本《説郛》看做三種抄本合成

的。見《説郛》（1927），跋，1a頁；《説郛三種》，1358c頁，及上海古籍出版社“説明”，載張宗祥“説郛校勘

記”，載《説郛三種》，1頁。從卷 67到 68，顯然改爲另手抄寫，但是這不一定意味着它們出自不同的抄

本。紙張的細微不同，與它們是相同紙張的不同印本也不矛盾。因此在這個問題上我傾向於張宗祥

的觀點。

③ 張宗祥跋中云“凡墨筆所書卷數有與目録不符者”，但是這肯定只是説目録與趙本、張本及其

他抄本内容有很小的差異。

④《中國古籍善本書目·叢部》，1.25a—38a（1990年重印版 49—75頁)；《中國古籍善本總目》，卷

六，集部（下），叢部，1923—1927頁。

⑤ 党寶海先生提出，“弘農楊氏”可能是指某個楊姓藏書家或書商。具體待考。

⑥ 印文：“杭州王氏九峰舊廬藏書之章”。 427


録》在卷55①。全目録已發表②。含序及（或）目録的卷已佚。編排與1927年商

務印書館版有許多不同。淺藍格，每半頁 9行，每行 34字。單魚尾，白口，全

書版心皆有卷號、頁號。

流傳史：明代抄本。有康熙時人（1662—1722）汪文柏（字季青，號柯庭，

安徽休宁人，占籍浙江桐鄉）印③。張宗祥 1952年發現於浙江圖書館，將其與

1927年商務印書館版對校，校勘記爲《説郛三種》編輯者收入附録④。

8. 滹本：現藏中國國家圖書館（no. A0485）⑤。δ系統。存 55卷，17册；不

含《聖武親征録》。目録、序皆不存，但一位晚近藏家在書皮的背面寫下了每

册内容目録。部分内容列在表 2中。藍格，每半頁 13行，每行 19—20字。白

口，雙魚尾，版心下書“滹南書舍”。四邊單框。

流傳史：可能是 1496年完成的郁文博本所據的一個範本。滹南書舍抄寫

於其抄書紙上。後來的一位藏家用作工作本（a working copy），有很多校對符

號（〇和丶），頁眉與行間有注⑥。這位藏家也在每册封面内頁加上了目録，顯

示出他的抄本目録已失。傅增湘有評⑦。

9. 史本：現藏中國國家圖書館，題爲“《説郛》不分卷”（no. A01507）⑧。ε系

統。12册，内容大約等於 20卷。無卷號⑨。内容列在表 4中。藍格，每半頁 14

行，每行皆 22字。頁面佈局與ε系統另一抄本沈本相同。白口，四周單邊，單

①《浙江圖書館古籍善本書目》所著録者，肯定與據稱含 45卷的抄本指的是同一抄本。後者見

《中國古籍善本書目·叢部》，1.38b頁（1990年重印版 76頁）；及《中國古籍善本總目》，卷六，集部（下），

叢部，1927頁。然而，現存卷的列表與浙江圖書館目録完全不合（見下）。

② 浙江圖書館古籍部編，《浙江圖書館古籍善本書目》，杭州：浙江教育出版社，2002，651—

654頁。

③ 印文：“休寧汪季青家藏書籍”、“古香樓”。

④ 上海古籍出版社“説明”，載張宗祥《説郛校勘記》，載《説郛三種》，1ff頁。

⑤《北京圖書館古籍善本書目·子部》，1694頁。參看《中國古籍善本書目·叢部》，1.38a頁（1990

年重印版75頁)；《中國古籍善本總目》，卷六，集部（下），叢部，1927頁。我在2009年7月對縮微膠捲做

了粗略考察，2012年7月考察得更久。

⑥ 我只得見黑白膠捲，無法確定這些批註是否如我們所期盼的那樣書以朱筆。

⑦《藏園訂補郘亭知見傳本書目》，752頁（所列的第二個《説郛》）。

⑧《北京圖書館古籍善本書目·子部》，1695頁；《中國古籍善本書目·叢部》，1.38b頁（1990年重印

本76頁）；《中國古籍善本總目》，卷六，集部（下），叢部，1927頁。

⑨ 後來的一位圖書管理者在一些地方加上了卷號。《大業雜記》標爲卷 57。但這不是數了史本的

真正卷數，而是取自γ本的卷號。428


魚尾，版心題“説郛”，無頁數①。

流傳史：明代抄本，優於嘉靖年間的沈本。有史夢蛟（乾隆時人）借樹山

房印②。眉批有三種：1）校正，見於前五册；2）卷號，取自γ本，粘貼或以筆書

寫；3）丁巳年曉鉦氏眉批，賈敬顔懷疑曉鉦爲錢大昕（字曉徵）、丁巳年爲嘉慶

二年（1798）。中國國家圖書館於民國時得之③。

10. 沈本：現藏香港大學馮平山圖書館（cat. no. 善 837/77-11）④。ε系統。

全 69卷，24册。卷 60爲《聖武親征録》。饒宗頤曾爲著録，附有代表性的卷的

内容⑤。無序、目録、卷號⑥。有陸樵跋四篇（見下）。白棉紙，黑格，每半頁 14

行，每行 22字。頁面佈局（mise en page）與ε系統的另一抄本史本相同。版心

三魚尾，有“沈”字題記。

流傳史：嘉靖抄本。據饒宗頤記，紙張與印章屬於沈瀚（字原約，吴江人，

1535 年進士）。後歸黄姬水（1509—1574）、陸樵，後者題跋云其得此書於己

丑，可能是 1589 年⑦。晚近曾歸吴江陸雲祥（字嘉卿，1627 年舉人）、盧址

（1725—1794）抱經樓、南潯劉承幹（字貞一，號翰怡，1881—1963），後歸馮平

山圖書館。

11. 臺本：現藏臺北“中央圖書館”（no. 000525628）。γ系統。全 100卷，64

册；卷 55 爲《聖武親征録》。全目録已發表⑧。有楊維楨序，卷頁署“都（郁之

① 我個人調查縮微膠捲；及《北京圖書館古籍善本書目·子部》，1695頁。

② 亦見賈敬顔，《聖武親征録校本》I，《綴言》，4a—b頁；《北京圖書館古籍善本書目·子部》，1695

頁。我所知的史夢蛟，僅以刊印全祖望（1705—1755）《鮚埼亭全集》《清全謝山先生祖望年譜》聞名。

③ 圖書館印文：“國立北平圖書館所藏”。

④ 見馮平山圖書館在綫目録：http://bamboo.lib.hku.hk/fpslindex/full_list.asp?RID+721。

⑤ Jao Tsong-yi,“Un inédit du Chouo-fou”, pp. 87—104；饒宗頤，《〈説郛〉新考——明嘉靖吴江沈

瀚鈔本〈説郛〉記略》，《饒宗頤史學論著選》，654—666頁；饒宗頤，《香港大學馮平山圖書館藏善本書

録》，158—164頁。

⑥ 饒宗頤在他的法文文章的一條腳註中提到（“Un inédit du Chouo-fou”, p. 90, n.1），實際上沈本

出現了四次卷號，但是數字與現今編次不一致。第 20册有一卷標爲六，第 21册有一卷標爲十五，第 22

册有一卷標爲四〇，第 24册有一卷標爲四〇。因爲沈本總共 69卷，所以看起來原來在後面的幾册被

移動到了前面。相似的現象也見於史本。

⑦ 雖然饒宗頤没有提供更多關於陸樵的信息，但鑒於該抄本之成肯定在 1535年之後，到乾隆年

間爲盧址所得，己丑年只有 1589 年、1649 年、1709 年三個可能性。跋中提及了最近的版本是 100 卷

本，而没有提及明末清初的120卷本，因此大概可以排除1649年、1709年，而1589年成爲唯一的可能。

⑧《“國立中央圖書館”善本書目》第2版，卷四，1445—1484頁，特别是1470頁。 429


訛）文博校正”①。内容與 1927年張宗祥的 100卷本、張本非常相似。每册有

單獨的目録。藍格，每半頁 11 行，無魚尾，版心無題記。尺寸爲 18.6 ×14.1

釐米。

流傳史：僞造品，1926年以後。《聖武親征録》文本複製自傅3本，並與王國

維1926年校注本對校過。這種對校，以及卷首加入郁（~都）文博，并删去卷97

的永樂時代作品，這些可能都是爲了造成版本價值高的假象。

12. 涵稿本：現藏浙江圖書館（no. 7437）。γ系統。全 100 卷皆存，52 册；

《聖武親征録》爲卷55。全目録已發表②。這個抄本是張宗祥1927年商務印書

館版的草稿，賈敬顔稱之爲涵芬樓本或涵本。序與内容與刻本一致。底本書

以墨筆，校正批注書以朱筆，校正通常來自其他抄本，但有時出於張宗祥的判

斷。因而，墨書的文本複製自張宗祥所用的諸抄本，即現存的傅本、張本、孫

本以及今佚的京本。因此，儘管多數卷僅僅有益於理解張宗祥的編輯過程，

但那些以今佚的京本爲底本的卷則是京本僅存的複製品。

今不知所蹤之抄本（皆爲γ系統抄本）

1. 京本：張宗祥 1927年版所用的一個抄本。（不可與傅—京本混淆，後者

賈敬顔稱爲京本。）張宗祥述此抄本存京師圖書館：

一爲京師圖書館殘卷（第三、第四、第二十三至第三十二）。無年

號。白緜紙。書極高大。似隆萬間寫本。③

京師圖書館藏品爲中國國家圖書館所繼承，所以推測此抄本應該仍然存世。

然而據我所知並没有任何目録著録之。如果張宗祥確實使用此抄本作爲他

的卷 3、4、23—32 的底本，那麽涵稿的這些卷應該就是大概準確地複製自京

本。更深入的譜系法分析，大概可以確定這個抄本在譜系中的位置。

2. 涉本：王國維述之，萬曆抄本，歸武進陶湘（字蘭泉，號涉園，1870—

1940）。王國維訪陶湘於天津，借之以校對其《聖武親征録》④。賈敬顔也提及

① 見昌彼得《説郛考》，圖一。

②《浙江圖書館古籍善本書目》，670—680頁。

③《説郛》（1927），跋，1a頁；《説郛三種》，1358c頁。

④ 王國維，《聖武親征録》，載《蒙古史料四種》，臺北：正中書局，1962，1b (2)頁。430


這部涉園本或稱涉本，但没有直接使用，他使用了 1901年日本重印的何秋濤

校、王國維注《聖武親征録》，藏於中國國家圖書館①。王國維、賈敬顔都没太

用這個版本，因爲它與傅 3本基本相同。確實，我對《聖武親征録》文本的對校

分析顯示這可能是個與傅 3本相同的複製抄本（codex descriptus）。既然王國

維强調了它與傅本的相似性，大概可以認爲它是γ系統的文本。

3. 傅—京本：賈敬顔述此抄本爲明代抄本，寫於“京師圖書館鈔”紙上，故

名之。因爲京師圖書館只在 1909—1928年存在，所以賈敬顔的意思肯定是，

它是明代抄本的現代複製本。賈敬顔稱此抄本現藏中國國家圖書館，但它作

爲一個 20世紀抄本顯然没有被收進國家圖書館善本書目②。我的譜系法研究

顯示，這個《聖武親征録》文本複製自傅 3本，並有一些校正，其中一些校正得

自當時最新出的《聖武親征録》版本——何秋濤校本。找到這個抄本有助於

理解張宗祥的《説郛》研究，但是對於《説郛》抄本研究而言並不重要。

4. 粤本：甲午年（1951）五月，張宗祥以廣州粤雅堂印書社抄本與其 1927

年版校對，後又以汪本校對。這兩次校勘的結果被《説郛三種》編輯者合併編

爲《校勘記》，但是編輯者未能區分與粤雅堂本校勘之記、與汪本校勘之記。

因此，這個《校勘記》中有來自兩個抄本的異文，以及張宗祥的校記。據我所

知，這個粤雅堂本迄今還没有找到。不過，既然汪本存世，以汪本與《説郛三

種》的《校勘記》做比較，就能排除汪本的異文，剩下的就是粤雅堂本的異文，

這可以用於尋找這個抄本。（也可能粤本就寫在粤雅堂紙上，這樣找它就更簡

單了。）《校勘記》包含卷 8、19、93，這些卷不見於存世汪本，至少從已刊佈的汪

本目録來看是這樣的。因此那些校勘記可能就構成了粤本。

表格

表1 《説郛》中北方、海外、邊疆主題作品的增删

α本= 1361年毛本；γ本= 趙本、鈕本、張本、臺北、1927年商務印書館版；ζ

本= 明末刻本（京都東方文化研究所藏本以及《説郛三種》重印本）

① 賈敬顔，《聖武親征録校本》I，《綴言》，5a頁。中國國家圖書館善本書目録中没有列這部書，大

概是因爲王國維批注的本子不是1894年版，而是1901年日本重印版。

② 賈敬顔，《聖武親征録校本》I，4a頁，5b頁。 431


主題

契丹遼

女真金

蒙元

高麗

越南

柬埔寨

緬甸

雲南

遼東

青海

回鶻

作者與時代

宋闕名

宋武珪

宋楊伯喦

宋王易

宋葉隆禮

五代胡嶠

宋洪皓（~邁）

宋文惟簡

宋闕名

宋程大昌

宋周煇

宋宇文懋昭

宋石茂良

宋孟（當作趙）珙

元闕名

元劉郁

宋孫穆

宋方鳳

宋松兢

元徐明善

元周達觀

唐闕名

元李京

元戚輔之

宋李遠

王延德

作品名

使遼録

燕北雜記

臆柬（~乘）

重編燕北録

遼志（節本）

陷虜記

松漠紀聞

虜廷事實

北風揚沙録

北邊備對

北轅録

金國志（節本）

避戎夜（~嘉）話

蒙韃備録

聖武親征録

西使記

雞林類事

夷俗考

使高麗録

安南行記

真臘風土記

驃國樂頌

雲南志略

遼東志略

青塘録

高昌行紀

α

48

40

9

—

—

—

29

3

49

—

—

—

—

—

—

—

31

—

—

—

—

—

—

—

—

—

γ

3

4

21

38

86

—

8

8

25

52

54

86

—

54

55

—

7

—

—

51

39

67

36

97

35

—

ζ

—

50

11

56

55

56

55

55

55

56

56

55

37

56

—

56

55

55

56

56

62

100

62

62

—

56432


材料來源：徐三見，《汲古阁藏明抄六十卷本〈说郛〉考述》，《东南文化》1994年第 6期，

118—127頁；中國國家圖書館，nos. 2408，3907；商務印書館編，張元濟序，《涵芬樓燼餘書

録》，上海：商務印書館，1951，卷三，子，57b—63a頁；《“國立中央圖書館”善本書目》第 2版，

卷四，1445—1484 頁；昌彼得，《説郛考》，第 2 版，臺北：文史哲出版社，1979，43—405 頁，

483—506頁；《説郛三種》，上海：上海古籍出版社，1988，10 卷；《東洋文化學院京都研究所

漢籍目録》，京都：東洋文化學院京都研究所，1938，324—347頁。

表2 《説郛》滹本與標準γ本之比較

滹本卷數

6

7

8

9

10

11

15

16

21

22

23

24

25

26

27

28

29

30

31

32

γ本中相應的卷數

3

4 前半

4 後半

5 前半

5 後半

6

8 前半

8 後半

11

12 前半

14 末

12 後半

13

14

15

16 前半

16 後半

17 後半

18 前半

18 後半

19 前半

所收作品

談壘（第二部分），古杭夢游録

墨娥漫録（第一部分，至仇池筆記）

墨娥漫録（第二部分，至封氏聞見記）

鶴林玉露

傳載

讀子隨識

玉澗雜書

捫蝨新話

玉泉子真録

悦生隨抄

博異志

洞天清録集

書鑒

就日録

因話録

三器圖義

雲林石譜, 宣和石譜

愛日齋叢鈔

坦齋筆衡

碧雞漫志

打馬圖經433


滹本卷數

33

34

35

36

37

38

39

40

41

42

64

65

66

67

68

69

70

71

72

73

74

75

76

77


19 後半

20 中

20 末

21 末

22 後半

23 前半

23-24

24 後半

75, 25

26

39 中

39 末, 40 首、末

41

43, 45

40 中

44

84

74 首、末

74 中

47

48

50

51

52

所收作品

甘澤謡

儒林公議

植跋簡談

昨夢録

山家清供

賓退録

諧史, 麈史, 歸田録, 孔氏雜說, 湘山

野録, 逸更諃 [訛]

墨客揮犀, 肯綮録

土林紀實, 卓異記, 集異記, 桐譜

宣政雜録, 洛陽名園記

陶朱新録,真蠟風土記

投轄録, 南窗紀談, 三楚新録,野説,

先公談録

宣室志, 驂鸞録, 攬轡録, 曲洧舊聞

宣靖妖化録, 錢氏私誌

慎子

靖康朝野僉言

錢譜

褚氏遺書, 大事記, 白虎通德論, 辨惑

論

大中遺事

公孫龍子

聱隅子歔欷瑣微論

識遺

豫章古今記, 安南行記

北邊備對

續表續表

434


滹本卷數

78

79

80

88

89

90

91


53 前半

53 後半

97

58

60

59 前半

59 後半

所收作品

鉤玄

四朝聞見雜録

金山志, 遼東志, 稽古定制, 勸善録,

夷堅志, 神僧傳, 效顰集

江表志

品茶要録

史記注語（第一部分）

史記注語（第二部分）

表3 《説郛》沈本與毛本、γ系統抄本之比較

沈本中的作品

學道玄真經

感應經

養魚經

相鶴經

相具（貝）經

土牛經

打馬圖經

酒經

讀北山酒經

醉鄉日月

品茶要録

宣和北苑貢茶録

北苑別録

大觀茶論

墨娥漫録

風土記……

仇池筆記

第一册

1卷

1卷

毛本

—

16

15

15

15

15

16

—

—

—

—

—

—

—

40

40

—

γ系統抄本

54

9

15

15

15

15

19

44

44

58

60

60

60

52

4

4

4

續表

435


沈本中的作品

(128 entries) 諸子隨識

文子……

尹文子……

淮南子……

論衡

諸傳摘玄

髙僧傳

無名公傳

蟹略

諸夷風俗

真臘風土記

第八册

卷？

第九册

卷？

末卷（第幾號?）

毛本

35

35

35

35

—

36

36

—

—

—

—

γ系統抄本

6

6

6

6

100

7

7

73

36

—

39

材料來源：饒宗頤，《〈説郛〉新考——明嘉靖吴江沈瀚鈔本〈説郛〉記略》，《饒宗頤史學

論著選》，上海：上海古籍出版社，1993，661頁；徐三見，《汲古阁藏明抄六十卷本〈说郛〉考

述》，《东南文化》1994年第 6期，118—127頁；昌彼得，《説郛考》，第 2版，臺北：文史哲出版

社，1979，43—405頁，483—506頁。

表4 《説郛》史本與γ系統抄本内容之比較

册

I.

II

III

書目

傳載

藏一話腴

墨客揮犀

續墨客揮犀

藝圃折中

讀子隨識

鬼谷子三卷

亢倉子

鬼谷子五卷

迷樓記

γ系統中的卷數

5

5

24

24

31

6

71

71

71

32

續表

436


册

IV

V

VI

VII

VIII

IX

書目

教坊記

卓異記

集異記

……柭印斗秤

趨朝事類

麟臺故事

北邊備對

省心詮要

甘澤謡

鐵圍山叢談

韋居聽輿

白獺髓

三水小牘

羣居解頤

鉤玄

稽古定制

遼志

遼東志略

金國志

雲南志略

煬帝開河記

……

墨子

子華子

曾子

尹文子

孔叢子

……萬機論

素書

聱隅子歔欷瑣微論


12

25

25

?

34

34

52

35

19

19

21

25

33

31

53

97

86

97

86

36

44

46

46

46

46

46

?

90

48

續表

437


册

X

XI

XII

書目

韓非子

北轅録

蒙韃備録

虜庭事實

溪蠻叢笑

……長城記

聖武親征録

大業雜記

嶺表録異記

海山記


47

54

54

8

5

?

55

57

34

32

《説郛》版本譜系圖

*表示用以校勘某一底本的參校本。

γ系统

趙本汪本* 鈕本* 張本*傅3本*涵稿本（紅墨）*

孫本* 傅3本涉本傅京本涵稿本（黑墨）

汪本史本沈本

續表

438

北京大學國際漢學家研修基地


Newsletter for

International China Studies

第九期2014. 6

圖書在版編目（CIP）數據

國際漢學研究通訊. 第 9期/北京大學國際漢學家研修基地編. —北京：北京

大學出版社，2014.9

ISBN 978-7-301-24670-2

I. ① 國… II. ① 北… III. ① 漢學—研究—世界—文集

IV. ① K207.8-53

中國版本圖書館CIP數據核字（2014）第190899號

書名：國際漢學研究通訊（第九期）

著作責任者：北京大學國際漢學家研修基地編

責任編輯：武芳翁雯婧

標準書號：ISBN 978-7-301-24670-2/K·1055

出版發行：北京大學出版社

地址：北京市海淀區成府路205號 100871

網址：http://www.pup.cn 新浪官方微博：@北京大學出版社

電子信箱：[email protected]

電話：郵購部 62752015 發行部 62750672

編輯部 62756694 出版部 62754962

印刷者：

經銷者：新華書店

720 毫米×1020 毫米 16 開本 33 印張 564 千字

2014 年 9 月第 1 版 2014 年 9 月第 1 次印刷

定價：82.00元——————————————————————————————————————

未經許可，不得以任何方式複製或抄襲本書之部分或全部内容。

版權所有，侵權必究

舉報電話：010-62752024 電子信箱：[email protected]

《國際漢學研究通訊》Newsletter for International China Studies

編輯委員會

主任榮新江

委員（按漢語拼音音序排名）：

白謙慎（美國波士頓大學）

程郁綴（北京大學）

程章燦（南京大學）

傅剛（北京大學）

寇致銘（澳大利亞新南威爾士大學）

李零（北京大學）

李慶（日本金澤大學）

劉玉才（北京大學）

馬辛民（北京大學出版社）

潘建國（北京大學）

齊東方（北京大學）

橋本秀美（北京大學）

榮新江（北京大學）

商偉（美國哥倫比亞大學）

王博（北京大學）

徐俊（中華書局）

楊繼東（美國斯坦福大學）

袁行霈（北京大學）

張志清（中國國家圖書館）

趙超（中國社會科學院）

鄭吉雄（香港教育學院）

主編劉玉才

目

録

目

録

漢學論壇

“關鍵字”、作者意圖和闡釋翻譯：司馬遷《報任安書》

康達維(David R. Knechtges) 撰施懿超譯/3

王肅《喪服要記》與漢魏時期的喪葬習俗廣瀨薰雄古橋紀宏/13

從洪邁《夷堅志》看宋代上下層文化的互動艾朗諾(Ronald Egan)/53

19世紀朝清古文家的交流初探

——以金邁淳與梅曾亮的交流爲討論的範圍金鎬/67

沈曾植與西本省三筆談考李慶/88

文獻天地

霞浦摩尼教文書《四寂讚》及其安息語原本

吉田豊撰馬小鶴譯/103

意大利漢籍的搜集高田時雄撰趙大瑩譯/122

美國哈佛燕京圖書館藏《永樂大典》十九庚形字史料價值辨析

辛德勇/134

斯坦福大學圖書館藏中文善本古籍簡介馬月華/154

《中國叢報》與中國古代文化文獻的翻譯張西平/163

《詩經》的東傳及其在朝鮮半島的刊行簡述蘇岑/180

琉球中央士族的漢籍校勘

——以楚南家文書爲中心水上雅晴/202 1


黎貴惇與《芸臺類語》趙培/216

漢學人物

饒公選堂之故事汪德邁撰李曉紅周軼倫房維良子譯/241

跨越語境新辟蹊徑

——謝柏軻與中國藝術史研究洪再新/252

對話馬悦然：“中國文學不必太否定自己”

——漢學家馬悦然教授學術專訪張玉梅/288

康達維專輯

美國當代漢學大師

——康達維（David R. Knechtges）先生的辭賦研究蘇瑞隆/302

正名·學統·知音：康達維對我的啓發及對美國漢學的影響

魏寧(Nicholas Morrow Williams)/320

博學審問取精用弘

——美國漢學家康達維教授的辭賦翻譯與研究馬銀琴/330

師門九載：側寫康達維先生之爲學、爲師與爲人連永君/343

康達維印象吴捷/353

Curriculum Vitae David R. Knechtges/357

Bibliography David R. Knechtges/378


《説郛》版本史

——《聖武親征録》版本譜系研究的初步成果

艾騖德(Christopher P. Atwood) 撰馬曉林譯/397

馬可·波羅與蒙古帝國的牌符党寶海/439

《馬可·波羅行紀》所載“妻女待客”風俗初探羅瑋/455

評《蒙元時代中國的東方敘利亞基督教》馬曉林/466

研究綜覽

“校勘與經典”國際學術研討會會議紀要唐田恬整理/481

2

目

録

“頻得音書似不遙”

——牛津大學“十至十三世紀中國精英的交流：

以書信與筆記作爲研究材料”國際學術工作坊

會議（2014.01.09—01.10）參加記王瑞來/486

基地紀事

“東亞漢籍研究：以日本古鈔本及五山版漢籍爲中心”

國際學術討論會會議綜述黃雅詩/497

國際漢學系列講座紀要（2013.10—2014.03） /502

徵稿啓事 /519

3

徵稿啓事

徵稿啓事

一、《國際漢學研究通訊》是北京大學國際漢學家研修基地主辦的綜合學

術刊物，辦刊宗旨爲報導國際漢學界在中國傳統人文學科領域的研究動態，

搭建中外學者溝通交流的學術平臺。本刊分設漢學論壇、文獻天地、漢學人

物、論著評介、研究綜覽、基地紀事等欄目，歡迎海内外學人賜稿或提供信息。

二、本刊暫定爲半年刊，分别在三月、九月底截稿。

三、本刊以中、英文爲主。來稿篇幅以中文一萬五千字内爲宜，特約稿件

不在此限。除經本刊同意，不接受已刊發稿件。論著評介欄目原則上不接受

外稿，但可以推薦。

四、來稿請提供Word文檔和PDF文檔，同時寄送打印紙本。中文稿件請

提供繁体字文本。如附有插圖，請提供原圖圖片格式（JPG 之類）的電子文

件。具体撰稿格式請參照文稿技術規範。因編輯人員有限，恕不退稿，請自

留底稿。咨詢稿件處理事宜，請儘量通過電子郵件。

五、來稿如涉及著作權、出版權方面事宜，請事先徵得原作者或出版者之書

面同意，本刊不負相關責任。本刊有權對來稿進行删改加工，如不願删改，請事

先注明。

六、來稿刊出之後，即致贈稿酬、樣刊。本刊享有已刊文稿的著作財産權

和數據加工、網絡傳播權，如僅同意以紙本形式發表，請在來稿中特别注明。

七、來稿請注明中英文姓名、工作單位、職稱，並附通信地址、郵政編碼、

電話傳真、電子郵件等項聯絡信息。

519


八、來稿請寄：

北京市海淀區頤和園路5號 100871

北京大學國際漢學家研修基地

《國際漢學研究通訊》編輯委員會

E-mail:[email protected]

附:

文稿技術規範

一、來稿請以Word文檔（正文五號字，1.5倍行距）打印紙本，同時提供電

子文檔。

二、來稿正文請按“一、（一）、1.、（1）”的序號設置層次，其中“1.”以下的

章節段落的標題不單獨占一行；文稿層次較少時可略去“（一）”這一層次；段

内分項的可用①②③等表示。

如：一、XXXX

（一）XXXX

1. XXXX

（1）XXXX。① XXX；② XXX；③ XXX。

三、來稿中的中文譯名，除衆所熟知的外國人名（如馬克思、愛因斯坦）、

地名（如巴黎、紐約）、論著名（如《聖經》、《資本論》）按照通用譯名外，其他人

名、地名、論著名在文中首次出現時，請括注外文原名，如沃爾特·福克斯

（Walter Fuchs)，地名、論著名照此處理。

四、來稿中的注釋，請採用頁下注、每頁各自編號，注號置於句末的標點

符號之前，如孔子已有“六藝”之説①，“……將邊界查明來奏”②。但引文前有

冒號者，句號在引號内，則注號置於引號之外，如《釋名》云：“經者，徑也，常典

也。”③

五、頁下注釋文字的具體格式如下：

1. 著作類：著作者名，《書名》，出版地：出版者，出版年（不加“年”字），X—

X頁。又：著作者名，《書名》卷X，X年X本。520

徵稿啓事

2. 雜誌類：著作者名，《論文名》，《期刊名》X年X期，X-X頁。又：著作者

名，《論文名》，《期刊名》X卷X號，X-X頁。

3. 西文書名與雜誌名均用斜體，文章名加引號。日文、韓文參考中文樣

式。

4. 重複出現的注釋不用“同上”簡略，但標注文獻出處只列著作、論文名

和頁碼即可。

例：① 郭紹虞，《宋詩話考》，北京：中華書局，1979，75頁。

② 張裕釗，《濂亭文集》卷四，清光緒八年查氏木漸齋刊本。

③ 袁行霈，《〈新編新注十三經〉芻議》，《北京大學學報》2009年2期，7頁。

④ 池田秀三著，金培懿譯，《韋昭之經學——尤以禮爲中心》，《中國文哲

研究通訊》第15卷3期，141-155頁。

⑤ Ad Dudink,“The Chinese Christian Books of the Former Beitang Library”,

Sino-Western Cultural Relations Journal XXVI (2004), pp. 46—59.

六、圖表按先後順序編號，在文中應有相應文字説明，如見圖X，見表X。

七、數字用法：

1. 公曆世紀、年代、年、月、日用阿拉伯數字，如18世紀50年代。

2. 中國清代和清代以前的歷史紀年、其他國家民族的非公曆紀年，用中

文數字表示，且正文首次出現時需用阿拉伯數字括注公曆。如秦文公四十四

年（公元前722），清咸豐十年（1860），日本慶應三年（1867）。

3. 中文古籍卷數均用中文數字表示，如作卷三四一，不作三百四十一。

521

1

The Textual History of Tao Zongyi’s

Shuofu: Preliminary Results of Stemmatic

Research on the Shengwu qinzheng lu1

By Christopher P. Atwood

(Indiana University)

Introduction Scholars of Song and Yuan-era literature have long been familiar with the Shuofu 説郛 or “Enclosure of

Literature” (SF), a vast anthology assembled by the private scholar Tao Zongyi 陶宗儀 (courtesy name

Jiucheng 九成 , sobriquet Nancun 南村 , 1316-1403)2 during the violent Yuan-Ming transition. First

compiled in 1361,3 the Shuofu was an example of the genre of anthologies (lèi shū 類書) which became

common in the Song dynasty (960-1279) as a way of dealing with the vastly increasing literary output in

China.4 After the founding of the Ming, Tao Zongyi 陶宗儀 continued to expand the SF. Committed to a

rather mild form of Yuan loyalism, his eclectic interests included a wide range of Inner Asian topics.5

1 My research on the SWQZL and the Shuofu has been aided by the kind assistance of many colleagues. I would like

particularly to thank Prof. Lucille Chia (University of California Davis), Prof. Dang Baohai 党宝海 (Peking

University), Ms. Wen-ling Liu (Indiana University, Herman B. Wells Library), Prof. Matsuda Koichi 松田孝一

(Osaka International University), Prof. Nakami Tatsuo 中見立夫 (Tokyo University of Foreign Studies), Prof.

Tachibana Makoto 橘誠 (Shimonoseki University), Prof. Ulaanbars (Qi Guang 齐光; Fudan University), Ms. Wang

Han 王函 (National Library of China), Dr. Wu Zhijian 吴志坚 (Hangzhou Library), Mr. Xu Sanjian 徐三见 (Linhai

City Museum), Mr. Zhou Qiao 周峤 (Fudan University), and Mr. Zhou Qing 周卿 (Shanghai Library), 2 Tao Zongyi’s dates have been a matter of controversy; I follow the conclusions of Chang Bide, Shuofu kao, pp.

407-482. 3 Until recently, one could only say that the earliest version of the Shuofu preceded Tao’s other great compilation,

the Nancun chuogeng lu 南村輟耕録, completed in 1366 (Chang Bide, Shuofu kao, pp. 12-13). The Mao 毛 (or

Jiguge 汲古閣) manuscript, however, preserves the date of the earliest draft as 1361; see Xu Sanjian 徐三见,

“Jiguge cang Mingchao liushijuan ben ‘Shuofu’ kao shu” 汲古阁藏明抄六十卷本《说郛》考述, Dongnan

wenhua 东南文化 1994, no. 6 (no. 106), p. 112.

4 On the leishu as a genre, see Wilkinson, Chinese History: A Manual, pp. 601-612, and Chinese History: A New

Manual, pp. 955-62. 5On Tao Zongyi’s life, see Chang Bide, Shuofu kao, pp. 2-10, 407-482, and Frederick W. Mote’s T’ao Tsung-i and

His Cho Keng Lu (Ph.D. Dissertation, University of Washington, 1954), pp. 1-12, 15-77, which is condensed in his

article, “Notes on the Life of T’ao Tsung-i,” Silver Jubilee Volume of the Zinbun Kagaku Kenkyusyo, Kyoto

2

Tao Zongyi was often criticized for credulity and superstitious beliefs, but his openness to the exotic, as

well as living under the Mongol Yuan dynasty, seems to have made him more aware of the broader world

than were most Chinese scholars. His 1366 collection of anecdotes, research notes, and common places,

the Nancun chuogeng lu 南村輟耕録 (“Nancun’s Notes Upon Rest from the Plow”) included a wide

range of information on the Mongols, semuren 色目人 (“peoples of various categories,” i.e. Westerners),

and other peoples in the Yuan.6 Among the rare and secret documents which Tao cited in this work was

the official genealogy of the Mongol imperial family, “The Genealogy of the Ten Ancestors” (Shizu shixi

lu 十祖世系録).7 His Shuofu incorporated both the Shengwu qinzheng lu 聖武親征錄, itself a lightly

edited version of the Veritable Records 實錄 of Chinggis Khan and Ögedei Qa’an, and the Meng-Da

beilu 蒙韃備錄, the only general description of the Mongols and Chinggis Khan written in his lifetime.8

In his Shushi huiyao 書史會要 (“Brief History of Calligraphy”), published in 1376, he included a chapter

on foreign scripts: Uyghur, Sanskrit, Japanese, and Arabic.9 In the Shuofu, he included not just but a

number of other works on Inner Asian dynasties and Southeast Asian kingdoms that otherwise might have

been lost (see Table 1).

During almost the entire Ming dynasty, the work circulated only in manuscript. Only about 250

years later was the work printed, in a blockprint edition carried out by the Wanweishantang 宛委山堂

University (Kyoto: Research Institute in Humanities of Kyoto University, 1954), pp. 279-293. Sun Zuo’s 孫作

biography written in 1374 is the single main source on Tao Zongyi; it is translated in Mote’s T’ao Tsung-i, pp. 29-31.

But Chang Bide has put together many isolated references to paint a much fuller picture.

6See the listing in Mote, T’ao Zongyi, pp. 147, 149-150, 160.

7 See YS 107/2729; cf. the note by Paul Pelliot in Hambis, Chapitre CVII de Yuan che, p. 144, and the discussion in

Christopher P. Atwood, “Six Pre-Chinggisid Genealogies in the Mongol Empire,” Archivum Eurasiae Medii Aevi,

vol. 19 (2012), 5-57. 8 The text is studied in Wang Guowei 王國維. “Meng-Da beilu jianzheng 蒙韃備錄箋證.” In Menggu shiliao

sizhong 蒙古史料四種. 1926: facsimile rpt. Taiwan: Cheng-chung Press, 1962, pp. 431-457 (1a-14b) + pp. 459-64. Translations include: Nikolai Tsyrendorzhievich Munkuev, trans., Men-da bei-lu: “Polnoe opisanie Mongolo-

Tatar”: faksimile ksilografa (Moskva, 1975); Peter Olbricht und Elisabeth Pinks, trans., based on the draft of Erich

Haenisch und Yao Ts`ung-wu, Meng-Ta pei-lu und Hei-Ta shih-lüeh: chinesische Gesandtenberichte über die

frühen Mongolen 1221 und 1237 (Wiesbaden, 1980); Möngkejayag-a, trans., in Ge. Asaraltu and Köke’öndör, ed.,

Bogda Bagatur bey-e-ber dayilagsan temdeglel (Höhhot, 1985), pp. 93-156. 9Mote, T’ao Zongyi, pp. 82-87, esp. 85-86.

3

publishing house of Hangzhou 杭州 during the Chongzhen 崇禎 era (1628-1644). This new block-print

eliminated some of the previous contents of the SF (including for the Shengwu qinzheng lu) and added

new works as well, expanding the 100 sections (juàn 卷) of most Ming-era SF manuscripts into 120

juàn.10 In the early Qing, the blocks were reused for several more reprintings, in which the contents were

rearranged and also preemptively expurgated of Song-era works that contained comments about

“barbarians” that the publishers thought might be offensive to their new Manchu sovereigns—the Meng-

Da beilu fell victim to this purge, for example.

Shuofu Studies to Date While the Shuofu includes a vast range of important materials found nowhere else, use of it has been

impeded by collection’s major textual problems. Both manuscripts and printed versions circulate in a

wide variety of versions differing radically in length and organization. Which version came first and what

are their interrelations are all questions on which eminent researchers such as Chang Bide 昌彼得 and Jao

Tsung-i 饒宗頤 have returned very different answers. Meanwhile, the progress of bibliographic

scholarship in the mainland of China has resulted in an increasing number of Ming-era manuscripts being

identified and catalogued. Yet how these manuscripts fit into the history of the Shuofu is still unclear.

As known to Qing-era scholarship, the Shuofu was a 120-juan work, with content focused mostly

on classical and literary topics—the Inner Asian and oversees exotica that formed a significant part of the

Ming-era manuscripts and the first 120-juan blockprint had mostly been purged. Catalogues however

occasionally noted the existence of Shuofus of various lengths, most often 100 juan, but also 60 or 70

10 There is a large literature about this printed edition of the SF, but many problems remain. See Chang Bide 昌彼得,

Shuofu kao 説郛考 (first edition 1962; revised and expanded edition Taipei: Wen-shih-che Publishing House, 1979).

The reprint of it in Shuofu sanzhong 説郛三種 (Shanghai: Shanghai Old Binding Press, 1988) gives the full contents

of original Chongzhen printing, but rather confusingly adds the Li Jiqi 李際期 and Wang Yinchang 王應昌 prefaces

which were first attached to the quite different early Qing 清 re-printing of 1646. Likewise the SF found in the

Kyoto Research Institute of Oriental Culture, whose contents are given in the catalogue Tōhō bunka gakuin Kyōtō

kenkyūsho kanseki mokuroku 東方文化學院京都研究所漢籍目録 (Kyoto: Tōhō bunka gakuin Kyōtō kenkyūsho,

1938), pp. 324-47, is confusingly said to date to Shunzhi 順治 3 (1646) and been sponsored by Li Jiqi, when it is

actually the Ming printing.

4

juan. Scholars such as Wang Guowei 王國維 and Paul Pelliot were especially intrigued by the evidence

that these Shuofu manuscripts, when accessed gave readings of works like the Zhou Daguan’s Zhenla

fengtuji 真臘風土記, the SWQZL and the MDBL that seemed far superior to the existing manuscript

traditions (which as it turns out in the beginning had all been derived from the Shuofu itself, although this

was not necessarily clear at the time).

Modern Shuofu studies began from their efforts in the 1920s, focusing on the relation of the late

Ming and early Qing printings to each other and to the 1496 mid-Ming version of Yu Wenbo 郁文博. Yu

Wenbo’s preface was found in the first printed Shuofu and it was often assumed (wrongly, as it turned

out), that his editorial activity must have been central to the manuscript tradition. This phase of research

was concluded by Chang Bide 昌彼得, in his Shuofu kao 説郛考 (first edition 1962; revised and

expanded edition, Taipei: Wen-shih-che Publishing House, 1979), which made basically obsolete

previous studies, such as those Kurata Junnosuke, King P’ei-yuan, Watanabe Kōzō, and Paul Pelliot.11

A landmark event in Shuofu studies was 1927 publication of a movable type version of the Shuofu

in 100 juan by Shanghai’s Commercial Press. Edited by Zhang Zongxiang 張宗祥, this edition was an

attempt to get behind the 120-section (juàn) blockprint edition and reconstruct in printed form the Shuofu

as it existed in manuscript before the 1620s.12 To do so, Zhang used four more or less fragmentary

manuscripts, all in 100 juàn but all incomplete, three of which contained the SWQZL, in chapter (juàn 卷)

11 Kurata Junnosuke 倉田涥之助, “Setsu-bu hanhon shosetsu to shaken”「說郛」版本諸說と私見, in Silver

Jubilee Volume of the Zinbun Kagaku Kenkyūsyo, Kyoto University (Kyoto: Kyoto University Humanities

Scholarship Research Center, 1950), pp. 287-304; King P’ei-yuan 景培元, Études comparative des diverses éditions

du Chouo fou, Scripta Sinica Monograph series (Beiping: Centre Franco-Chinois d’Études Sinologiques, 1946),

Watanabe Kōzō 渡邊幸三, “Setsu-bu kô 説郛考,” Tôhô gakuhô 東方學報 (Kyoto) 9 (1938), pp. 218-60; and Paul

Pelliot, “Quelques remarques sur le Chouo fou,” T’oung Pao 23 (1924), 163-220. 12Tao Zongyi, Shuofu (Shanghai: Commercial Press, 1927), 100 juàn in 40 volumes in four cases, described as “a

typeset edition collated on the basis of Ming-era MSS preserved in the Wetlands Fragrance House (Hanfenlou 涵芬

樓).” In Chinese, this is generally known as the Hanfenlou 100-juàn edition. This edition was reprinted by the

Commercial Press in Taipei in 1972, and also in 1988 as the first two volumes of the ten-volume set Shuofu

sanzhong. Volume 10 of this edition has an index to the works; Chang Bide, Shuofu kao, pp. 43-405, 483-506

provides both an index and a brief description of all the works found in this Zhang Zongxiang edition.

5

55. One of these two was a set of fragmentary SF mss acquired by Fu Zengxiang that Wang Guowei had

used to establish the “Shuofu text” of the SWQZL, whose parts variously dated to the late fifteenth century

and 1505,13

and the other was a Wanli era (1573-1619) ms kept by the Commercial Press in its “Wetlands

Fragrance House” (Hanfenlou 涵芬樓) in Shanghai.14 A third one was a ms kept by the scholar Sun

Yirang, of which Zhang Zongxiang had a copy made.15 Scholars of the Shuofu text were disappointed that

Zhang collated the mss without any scholarly apparatus and took aim at Zhang’s overly ambitious claim

to have reconstructed Tao Zongyi’s original Shuofu. Thus, Watanabe Kōzō and others early on pointed

out the presence of texts of the Yongle era (1403-1424) that refuted Zhang’s claim to have reconstructed

Tao’s original form as created in the Hongwu ere (1368-1399).16 In reality, all the manuscripts used by

Zhang Zongxiang dated from after 1450 and contained texts that could only have been added to the text

after Tao Zongyi’s death. Yet even so, this new version was far closer to Tao’s original work than the

block-printed 120-juan Shuofu.

Since the publication of Chang Bide’s research, studies outside of China came to a long stand still.

Only two Ming-era manuscripts of the Shuofu exist outside of mainland China (one in Hong Kong and

one in Taiwan), so it was not until the resumption of scholarly activity in the People’s Republic in the late

1970s, that Shuofu studies began to slowly recommence. Since the 725-title, 100-juan manuscript

tradition seemed to be adequately represented by Zhang Zongxiang’s edition, research has tended to

focused on the identification and description of mss of the Shuofu independent of the 100-juàn manuscript

tradition and the relation of 100-juàn manuscripts to the original Shuofu of Tao Zongyi. Given the

tremendous bulk of the Shuofu manuscripts, essentially all of which are missing at least a few juan,

scholars have mostly focused on trying to match the tables of contents with the cryptic suggestions in

Ming scholarly writings that suggested the existence of earlier, non-100 juan Shuofus.

13 I follow Jia in designating this composite set the Fu ms; see more in the Appendix. 14 I follow Jia Jingyan in designating this as the Zhang ms 張本 (Zg); see more on it in the appendix.

15 I designate this the Sun 孫本 (Sn) ms; see more on it in the appendix. 16 Watanabe, Setsu-bu kô, p. 230; King P’ei-Yuan, Etude comparative, pp. 3-4.

6

The Stemmatic Approach Meanwhile a whole different line of approach was being pioneered by the late Jia Jingyan 賈敬顔

(courtesy name Baiyan 伯顔, 1924-1990), professor of history at Minzu University in Beijing. Working

on a critical edition of the SWQZL which he completed and printed in mimeograph form in 1979, he

identified several manuscripts of the SF that contained the SWQZL and by detailed comparison of their

texts arranged these manuscripts in a tentative order of least corrupt to most corrupt. He also gave all the

SF manuscripts he worked with convenient names and descriptions that summarized much of the

emerging catalogues of rare manuscripts in China. Had it achieved wide circulation, the relevance of this

work to the study of the Shuofu would have been immediately apparent, but due to being printed only in

mimeograph form, Jia’s edition of the SWQZL did not achieve due recognition or wide distribution,

outside of a few Mongolian studies scholars.

Jia’s research showed that the interconnections between various SF manuscripts could be studied

not just by the extensive survey of their contents and arrangement of works, but also by the intensive

study of one (or more) selected works contained within the SF. As is well known in the stemmatic

methodology of textual criticism, examination of works copied in manuscript can identify how each

manuscript inherits certain indicative errors from the exemplar or exemplars from which it was copied,

and in turn adds a few more such errors which it passes down to all manuscript copied from it, and so on.

Thus careful examination of multiple manuscripts permits the researcher to draw up a “family tree” or

stemma, which indicates the relationships between the manuscripts examined. Although Jia did not use a

stemmatic methodology his worked opened the way to doing so, by locating and giving initial

descriptions of most of the relevant manuscripts currently available in libraries.

The greatest practical difficulty in drawing up such a stemma is determining in any given place

which is the primitive reading and which is the derived one. Particularly when both readings make some

kind of sense, such a determination is often frustratingly subjective. It is here that Tao Zongyi’s interest in

exotica from the defunct Yuan dynasty gives a crucial advantage. The SWQZL in particular is perfectly

7

suited to such an analysis because it has both a complete Persian parallel and a partial Mongolian parallel.

Moreover, the extensive Mongolian transcriptions are such that random corruptions in the Chinese

manuscripts can usually be immediately detected because they result in names which are not

reconstructable as Mongolian.

Two examples will show the utility of these controls:

In a name which some manuscripts consistently give as Beilu Kehan 盃禄·可汗, the first

character bēi 盃 is sometimes found as bēi 杯, mèng 孟, yíng 盈, or mì 覔. Comparison with Mongolian

histories shows, however, that this name corresponds to that of Buyruq Qa’an, and that while bēi 杯

might be a possibly primitive reading, mèng 孟, yíng 盈, or mì 覔 cannot be correct, and those readings

must be the result of textual corruption.

In a second example, under the autumn of year gui/you 癸酉, there is a description of a Mongol

siege of the city of Zhuozhou 涿州. Some manuscripts say the city fell, within the “specified time” (keri

刻日), others that two previously specified commanders were “both commanded” (er ming ri 二命日) to

take it, another has a strange reference to a possible divination (er bu ming ri 二卜命日), while the two

others say the siege took more than twenty days (ershiyu ri 二十餘日). While some of these readings

might be more acceptable than others, a final decision would be difficult, except that the Persian parallel

of this passage, in the Compendium of Chronicles by Rashīd al-Dīn, states clearly that the Mongol armies

“laid siege for twenty days and captured the city.”17 Thus it is the last version which is unquestionably

primitive and the others all show a greater or lesser degree of corruption. In other passages, the parallel

offered by the Secret History of the Mongols, which the Shengwu qinzheng lu cited extensively as its

source, can also assist in determining which reading is primitive and which is derived. Assembling such

17 Rashid ad-Din, trans. O.I. Smirnova, notes by B.I. Pankratov and O.I. Smirnova, ed. A.A. Semenov, Sbornik

letopisei, Vol. I, part 2 (Moscow: Academy of Sciences Press, 1952), p. 169; Rashiduddin Fazlullah, trans. and ed.

W.M. Thackston, Jami‘u’t-Tawarikh: Compendium of Chronicles: A History of the Mongols (Cambridge: Harvard

University Press, 1998), vol. 1, p. 219.

8

bits of evidence thus makes a clear stemma or family tree, showing the relationship of the SWQZL texts

within the Shuofu manuscripts. Such a relationship can with caution then be treated as a preliminary

hypothesis for the stemma of the SF manuscripts as a whole.

As with Jia Jingyan, Wang Guowei, and Paul Pelliot, my interest in the text of the Shuofu is the

outcome of my previous interest in the SWQZL. As part of preparing a critical edition of the Shuofu, I

have collated the text of the SWQZL found in nine different copies of the Shuofu, and have also examined

all the readings supplied by previous scholars for the SWQZL found in three different Shuofus whose

location is currently unknown. These nine include all but two of the major Shuofu manuscripts listed in

the major Chinese catalogues of rare books and/or discussed in the literature on the Shuofu. They are as

follows (I have adopted here the convenient names for them given by Jia Jingyan), listed with their

current location in rough order of most primitive to most derived:

Zhao 趙: National Library of China, Beijing

Niu 鈕: National Library of China, Beijing

Sun 孫: Yuhailou 玉海樓 museum, Ruian 瑞安, Zhejiang

Zhang 張: National Library of China, Beijing

Fu3 傅(3)18: Shanghai Library

Taipei 台: Central National Library, Taipei

Uang 汪: Zhejiang Library, Hangzhou

Shi 史: National Library of China, Beijing

Shen 沈: Fung Ping Shan Library, University of Hong Kong.

18 The Fu ms is a composite ms, comprised of three or four different fragmentary Shuofus, boxed together to make

an almost complete set. The SWQZL is in the third part, which I thus designated Fu3.

9

The three manuscripts which were collated by previous scholars, but whose location is currently

unknown are:

Chang 閶: Copy made in Suzhou and used by Zheng Jie 鄭杰 in his unpublished 1778 study of

the SWQZL

She 涉: used by Wang Guowei for his 1926 edition of the SWQZL; Wang’s notes also used by Jia

Jingyan for his SWQZL

Fu-Metropolitan 傅京(師): used by Zhang Zongxiang for the 1927 Commercial Press edition of

the Shuofu, and by Jia Jingyan for his SWQZL. So-called because it is a copy of the Fu ms made

on stationary of the Metropolitan Library in Beijing.

Further information on these mss will be found in the appendix.

My stemmatic research has led to a number of important conclusions, for example that the Shen

ms in Hong Kong is not close to Tao Zongyi’s original one, as was suggested by its preface and by Jao

Tsung-i, but is actually a rather late and corrupt mid-Ming version, and that the Taipei ms, upon which

Chang Bide based his research, is actually a twentieth-century forgery.

I have also taken the opportunity to examine the other two manuscripts which do not contain the

SWQZL, that is the Mao 毛 or Jiguge 汲古閣 ms in the Linhai City Museum (Linhai shi bowuguan 臨海

市博物館) and the Hūnan Printing House (Hunan shushe 滹南書舍) ms, kept in the National Library of

China. While these ms do not contain the SWQZL and thus cannot be directly added to the stemma, they

are very distinctive in organization and I believe they can be provisionally related to the picture of the

Shuofu’s development set out here.

The currently extant Shuofu texts (including the first blockprinted version) can be divided into

five different recensions, each differentiated on the basis of length and/or organization. To these may be

added another, unfortunately non-extant, recension whose basic organization can be surmised from the

10

internal evidence of two other recensions. Following usual text-critical practice, I label them with Greek

letters, listed here with the exact or approximate date of completion of its earliest exemplar and extant

exemplars:

Α: 60 juan, 366 titles; dated to 1361; extant in Mao ms.

Β: 100 juan, c. 600 titles; c. 1370; not extant, but contents roughly reconstructable from γ

Γ: 100 juan, 725 titles; c. 1440; extant in Zhao, Niu, Sun, Zhang, Fu, Taipei, and Uang mss, and

the 1927 Commercial Press printed edition

Δ: 100 juan, c. 650 titles; 1496?; extant in Hu ms

Ε: 69 unnumbered juan, estimated 725 titles; Jiajing era (1521-1566)?; extant in Shi and Shen

mss.

Ζ: 120 juan; 1,236 titles; c. 1615; extant in the Wanweishan Tang blockprints

In the rest of this article I will survey what is known of these recensions, how they were created, and their

interrelations.

The Α Recension In its original form, the Shuofu consisted of 60 juàn as described in the preface written by Tao Zongyi’s

friend and Yuan loyalist Yang Weizhen:

Master Tao Jiucheng of Tiantai 天台 has taken books from the classics and histories and

biographical narratives on down through the varied writings of the hundred schools, works of

more than a thousand authors, and has compiled them into 60 juàn 卷 totaling many tens of

thousands of passages. He has named it Shuofu 説郛, taking part of a sentence from the great

11

thinker Yang 楊.19 He asked me to write a preface for it. I have read it over the space of months.

It has been able to supply facts which my studies have overlooked. Scholars obtaining this book

will find that it can expand to a great degree what they have heard and seen.20

The preface was dated to two days before full on the ninth moon of autumn, year xīn/chǒu 辛丑 of the

Zhizheng 至正 period, or November 12, 1361.21 Fifteen years later, in a preface to Tao’s “Brief History of

Calligraphy,” another of Tao’s distinguished friends, the famous early Ming Confucian and chief editor of

the Yuan shi, Song Lian 宋濂, described the compilation thus:

Jiucheng has read various biographical narratives by more than a thousand authors, most of which

are ones the world of scholarship has never seen. So he imitated Zeng Zao’s 曾慥 Lèishuō 類説

(“Classified Literature”)22 and made the Shuofu (“Enclosure of Literature”) in a certain number of

chapters (juàn 卷). Whenever there was something he compiled, he abridged it; gentlemen called

the resulting words both very deep and very broad.23

In other words, the work was basically a set of Tao Zongyi’s “reading notes” (dúshū bǐjì 讀書筆記), and

as such bore the imprint of his eclectic character.

19Yang Xiong 楊雄 (53 BC-AD 18) Fayan 法言, juàn 4 reads: “Within the borders of Heaven and Earth all things

are enclosed; within the limits of the Five Classics all other literature is enclosed.” Thus the anthology’s name

“Enclosure of Literature” implied that it contained a range of works, yet all contained within the bounds of the Five

Classics. See Mote, T’ao Tsung-i and His Cho Keng Lu, pp. 79, 100; Paul Pelliot, “Quelques remarques,” 163 n.1.

20Text in Shuofu (Shanghai: Commercial Press, 1927), “Shuofu xu [2nd],” p. 1a; Shuofu sanzhong (Shanghai:

Shanghai Antique Binding Press, 1988), vol. 3, pp. 1; Mote, T’ao Tsung-i, p. 79. I have replaced the “100 juàn”

found in all later copies and editions with the original “60 juàn” as found in the Mao ms; see Xu Sanjian, “Jiguge

cang Mingchao liushijuan ben,” pp. 112, 116.

21The date is found only in the Mao ms; see Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 112, 116.

22The Leishuo 類説 of Zeng Zao 曾慥 (1091-1155) is one of the earliest anthologies. Zeng Zao also composed an

anthology of Daoist texts, the Daoshu 道樞 (“Pivot of the Dao”) in 108 chapters.

23Cited in Chang Bide, Shuofu kao, p. 10-11 and Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” p. 117; partial

English translation in Mote, T’ao Tsung-i, p. 100.

12

The only manuscript of the Shuofu that preserves its original 1361 form is the 60-juàn Mao ms,24

now kept in the Linhai City Museum, Zhejiang province. It was described in 1994 by Xu Sanjian, with a

full table of contents.25

It stands out for three very distinctive characteristics: its sloppiness, its brevity,

and its organization. The writing was described by one collector, Mao Yi 毛扆 in 1710, as having

“blunders cropping up everywhere” and “almost unreadable” and by the twentieth-century bibliophile and

scholar Zhang Zongxiang 張宗祥 (1882-1965) thus: “It was copied only by a vulgar hand, is dotted with

wrong characters, and almost unreadable.”26 The tendency to use alternate characters (tōngjiǎzì 通假字),

often based on the author’s native dialect, is pervasive, make er 兒 into li 立, jue 覺 into jiao 角, and zhi

治 into huo 活, and so on.27 Despite these errors, however, the Mao ms has already demonstrated its great

value for textual research.28 It is also the shortest known complete Shuofu text, containing only 60 juàn

and 366 separate titles.

The Mao ms is also the only Shuofu ms with a consistent organization. Rather than the topical

organization of other anthologies (lèishū 類書), however, the extracts are classified according to the final

24 I give it this designation from being held by Mao Jin 毛晉 (1599-1659) in his famous Jiguge 汲古閣 (“Chamber

for Exploration of the Classics”) Library. Xu Sanjian calls it the Jiguge ms, but in line with Jia’s practice of using

single-character names based on the ms’s earliest or best-known possessor, I prefer to call it the Mao ms. The

manuscript’s provenance is given in Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” p. 112. Mao Jin wrote a

colophon to the Nancun Chuogenglu referring to the 100-juàn Shuofu as an incomplete work; see Tao Zongyi 陶宗

儀, Nancun chuogenglu 南村輟耕録 (Beijing: Zhonghua Shuju, 1997), p. 385; evidently he was judging from the

contrast between the 100 juàn widely referred to and the 60 juàn in the copy in his possession.

25Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 112-27. There is also a good description of this ms, with

photographs of selected pages in the chapter “Shanben miji lun ‘Shuofu’ 善本秘籍《说郛》” of Zhou Xiangchao

周向潮 and Xu Sanjian 徐三见, ed. Lishi wenhua mingcheng Linhai 历史文化名城临海 (Hangzhou: Zhejiang

People’s Press, 2002), pp. 245-48.

26Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 113, 115.

27Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 116-17. On tōngjiǎzì 通假字, see Wilkinson, Chinese

History, pp. 421-23, and Chinese History: A New Manual, pp. 45-46.

28 See Wu Jing 吴晶, “‘Shuofu’ ben ‘Luoyang qielan ji’ de banben jiazhi” 《说郛》本《洛阳伽蓝记》的版本价

值, Nanjing shifan daxue wenxueyuan xuebao 南京师范大学文学院学报, Mar. 2009, no.1, pp. 15-17. Wu notes

that despite some obvious errors like 蒙 for 萊, the Mao ms text of the Luoyang qielan ji: 1) preserves correct

readings and omitted sentences found in no other ms or edition; 2) helps decide between readings found in two large

classes of sources; and 3) demonstrates superior readings compared to those found in the 1927 Commercial Press

edition of the Shuofu, based on γ recension manuscripts.

13

character(s) in the works’ titles. The final part of the title was usually a word such as “notes” or “records”

or “biography,” so his method amounted to a rather crude arrangement by genre. Juàn 1-14 was a special

section for works already collected into cóngshū 叢書 (“collection”), 15-16 for those ending in jīng 經

(“classic”), 17-18 for shǐ 史(“history”), 19-20 for biān 編 (“compilation”), followed by pǔ 譜 (“register”),

chāo 抄 (“copy”), bǐjì 筆記 (“notes [on readings or lectures]”), jìwén 紀聞 (“notes on contemporary

events”), tán 談 (“discussion”), shì 事 (“narratives”), huà 話 (“talk”), shuō 説 (“tales”), zhì 志 (“treatise”),

jì 記 (“memoirs”), and finally lù 録 (“records”).29 The last section would have been where the SWQZL

would be but like many other works known to have been in the Shuofu later, it is not found in the Mao ms.

Clearly this late Yuan manuscript was not the final version of the Shuofu.

The Β and Γ Recensions At present, every other known manuscript of the Shuofu apart from the Mao ms is based at least partially

on exemplars deriving from a later class of Shuofu manuscripts with 725 titles and organized into 100

juàn—what I call the “γ recension.” Manuscripts of the γ recension, which are quite the most common

type of Ming manuscript of the Shuofu, all have a similar organization and contents, quite different from

the α recension’s Mao ms. Comparing the contents of the Mao ms, as representative of the α recension,

and the various exemplars of the γ recension, all the works found in the α recension’s 60 juàn are

crowded into the first 30 juàn of the γ recension. Thus 70 additional juàn of new material were then added

to the γ recension, raising the total number of works included from 366 to about 725.30 The old

29Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” p. 113, and the table on pp. 118-27.

30 In a detailed comparison of the contents of the Mao ms Shuofu and the γ recension, as seen in Zhang Zongxiang’s

printed 100-juàn Shuofu, I found that only 6 of the 366 Mao ms Shuofu titles and sub-titles were found in any juàn

past 30 in the γ recension. Likewise out of the 281 titles found in the first 30 juàn of the γ recension, only 36 (not

counting duplicates and works taken from the Baichuan xuehai) do not derive from the Mao ms. (The difference in

count of titles comes from differences in whether the extracts in topical anthologies are counted separately or only

under the anthology’s larger title.) Published contents of other γ recension mss, all essentially identical to that of

Zhang Zongxiang’s, include that of the Zg ms in Shangwu yinshuguan, ed., Hanfenlou jinyu shulu (Shanghai:

Commercial Press, 1951), vol. 3, Zi 子, 57b-63a, and the T ms in Guoli zhongyang tushuguan shanben shumu, 2nd

edition, vol. 4, pp. 1445-84.

14

organization according to the last character of the title was mostly ignored in the new material, and even

the titles inherited from the old Shuofu in the first thirty juàn were shuffled around. As a result, the

material in the γ recension was organized according to a confusing mix of topical, genre, and final

character considerations.31 These γ recension exemplars still have the old Yang Weizhen preface, but its

date was removed and the count of 60 juàn was changed to 100 juàn to fit the new scale of the anthology.

One way in which the γ recension mss resembled the old α recension was in the frequency of

simplified and unorthodox characters. These can be seen particular in the Zhao, Zhang, and Fu3 mss

which on stemmatic evidence are relatively primitive and which retain a large number of simplified

characters. Although the kind of tōngjiǎzì 通假字 and outright errors found in the Mao ms are not

common, simplified forms used commonly include bèi 俻 (for 偹), chēng 称 (for 稱), gūi 帰, nán 难, qín

禽 (for 擒 “capture,” not “animal”), and tīng 听 while others used inconsistently include bào 报, biān 边,

fŭ 抚, huán 还, tān 摊, wú 无, suǒ 𠩄, yīn 囙, hào 号, jǐn 尽, jù 惧, shì 執 vs. zhí 执, sūi 虽, yǔ 与, and

zǒng 捴. Just as distinctive as the use of these vulgar forms is their inconsistency, even within a single

text such as the SWQZL.

All but one of the extant γ recension mss date to the sixteenth century or after. The Zhao ms is

dated to year gēng/shēn 庚申 of the Hongzhi era (i.e. 1500) and the Fu3 ms (that is the third of three or

four fragmentary Shuofu texts together forming an almost complete Shuofu first described by twentieth-

century scholar Fu Zengxiang 傅增湘) is dated to Hongzhi 弘治 18 (1505).32 Almost certainly earlier

than either of these, however is the first of the three or four mss in Fu Zengxiang’s Shuofu, containing

juàn 1-25. This manuscript, which I designate Fu1, was written on paper of the Congshutang 叢書堂

31 In the Mao ms, for example, Tao Zongyi began with 14 juàn of cóngshū or collectanea—smaller collections now

to be included in a larger one. In the γ recension, the material on the Confucian canons (jing), previously in juàn 34-

35 was given pride of place in juàn 1 and 2. Later, an anonymous editor reorganized the 100-juàn Shuofu in a new

way (the ε recension, exemplified by the Shen and Shi mss), putting all works concluding in the character jing 経,

many of which were recently written canons of taste (wine, horse-riding, etc.), not real classics, in juàn 1.

32 See the manuscript descriptions on these four Fu mss. Based on my examination of the ms in the Shanghai Library,

the Fu1 ms does not have a preface.

15

library of Wu Kuan 吳寬 (1435-1504). 33 Jao Tsung-i guessed that it might date to a few years earlier than

Chenghua 成化 8 (1472), when Wu Kuan received his jinshi 進十 degree.34 Since all the Fu mss have the

same organization (which is what enabled them to be cobbled together into a single almost complete

Shuofu) this pushes the date of the earliest γ recension back to the third quarter of the fifteenth century.

Moreover the text of the SWQZL in the Zhao and Fu3 mss are sufficiently different that several instances

of copying must have elapsed between their time and that of their common ancestor, again pushing the

date of their common ancestor well back into the fifteenth century. Thus, the expansion of the 60-juàn

Mao ms into the γ recension was an affair of the mid-fifteenth century at the latest.

Some of the works added to this γ recension, including the SWQZL, appear to have been copied

relatively early in the Yuan-Ming transition, when Tao Zongyi still thought of himself as Yuan loyalist.

To each work, Tao added the author (where known) and the author’s dynasty. Several works of the Yuan

dynasty appear in the γ recension with the authors dated to the Huang Yuan 皇元 “Sovereign Yuan”

dynasty. And two works have the author dated to the “End of the Song, beginning of our dynasty” (Song

mo guo chu 宋末國初). 35 In the text of the SWQZL, references to the Mongol emperors are always given

special honorific spacing, a feature probably of the original work, but once scrupulously preserved by Tao

in his copying.

33 On the fragmentary mss put together to form the Shuofu of Fu Zengxiang 傅增湘 (1872-1950), the best source is

Mo Youzhi 莫友芝, supplemented by Fu Zengxiang 傅增湘, ed. Fu Xinian 傅熹年, Cangyuan dingbu Lüting

zhijian zhuanben shumu 藏園订補郘亭知見傳本書目 (Beijing: Zhonghua shuju, 1993), vol. 2, 10B/751-52. The

descriptions given by Zhang Zongxiang in his colophon to his 1927 edition (Shuofu [1927], colophon/1a; Shuofu

sanzhong, p. 1358c) is least accurate; somewhat better is the 1988 “explanation” (shuoming 説明) in Shanghai guji

chubanshe 上海古籍出版社, “Shuoming 説明” in Zhang Zongxiang 張宗祥, “Shuofu jiaokan ji 説郛校勘記” in

Shuofu sanzhong 説郛三種 (Shanghai Old Binding Press, 1988), p. 1. 34 See Jao Tsong-yi, “Un inédit du Chouo-fou: Le Manuscript de Chen Han de la période kia-tsing (1522-1566),” p.

93. 35 These dates were first remarked on by King P’ei-yuan, p. 4. Huang Yuan 皇元 appears in the following works

(references are to juàn and page number in the printed 1927 Shuofu): Chun meng lu 春夢録 (42/18b), Annan xingji

安南行記 (51/18b); Shengwu qinzheng lu 聖武親征録 (55/1a), Anya tang jiuling 安雅堂酒令 (56/1a), Jingbei yin ji

鯨背吟集 (57/1a). Song mo guo chu 宋末國初 appears in Gusu biji 故蘇筆記 (57/20a) and Xue zhou cuo yu 雪舟脞

語 (57/20b).

16

Judging from these features, all of these works should have been added to the Shuofu while Tao

still considered himself a Yuan man. Together these include all the works attributed to the Yuan between

the juàn 40 to 60, except for Kunxuezhai zalu 困學齋雜録 whose author is dated merely to the Yuan 元.36

Yuan-era works from juàn 64 on, however, are dated simply to the Yuan 元. Oddly, though, those works

which precede juàn 40, including many copied into the Shuofu already in the Mao ms, simply have Yuan

元, even though Tao Zongyi was then certainly writing under the Yuan dynasty.37 My guess is then that

the materials in juàn 42 to 57 with the Huang Yuan attribution were added during a period when Tao was

particularly concerned to emphasize his Yuan identity, most likely from the time of his sister and sister-

in-law’s deaths in 1367 to the first year or two of the new Ming dynasty. This should be the period when

he was collected the works that later formed juàn 40-60 of the SF.

Other titles, however, were certainly added to the Shuofu much later. All of the γ recension mss

contain several Ming dynasty works:

1. Qian pu 錢譜, described as an anonymous of the Ming, containing references to the Yongle era

(1402-1424); in juàn 84

2. Gu ge lun 古格論, by the Ming author Cao Zhao 曹昭, and dated to Hongwu 洪武 21 (1388); in

juàn 87

3. Quan shan lu 勸善録, by the Ming Empress Renxiao Huanghou, maiden name Xu 徐 (1362-1407,

enthroned as empress 1403); in juàn 97;

36 See 1927 Shuofu 52/17a. 37 See Pei chu xuan ketan 佩楚軒客談 (7/22b; Mao ms, juàn 11), Hua jian 畫鑒 (13/1a; Mao ms, juàn 6); Suichang

shan qiao zalu 遂昌山樵雜録 (19/6; Mao ms, juàn 58), Haoranzhai yi chao 浩然齋意抄 and Haoranzhai shiting

chao 浩然齋視聽鈔 (20/1a and 7a; juàn 24), Shanfang suibi 山房隨筆 (27/12a; Mao ms, juàn 28). In some cases

there is conflict over the era of the author. For example Qiantang yishi 錢塘遺事 (7/29a; Mao ms juàn 31) is

attributed to the Song in the Mao ms, but to the Yuan in the 100-juàn Shuofu mss. I have included only those

attributed to the Yuan in both the Mao ms and the 100-juàn Shuofu mss.

17

4. Xiao pin ji 效顰集, by the Ming author Zhao Bi, with an early draft in middle Yongle 永樂 to

Xuāndé 宣德 3 (1427) and probably completed in Zhengtong 正統 1 (1436); in juàn 97.38

By the time of these last two works, Tao Zongyi was certainly dead; thus the γ recension as attested in the

existing mss was certainly completed not by Tao Zongyi, but by a continuator or continuators. There is

one possible clue to the identity of this person. In the Fu1 ms, which appears to be the earliest surviving

exemplar of the γ recension, there is a statement about editorial activity. After a statement that it was

compiled by Tao Zongyi (Nancun zhenyi Tao Zongyi zuan 南村真逸陶宗儀纂), there is another line

stating that it was edited by Gong Fu of Nanzhai (Nanzhai Gong Fu jiaozheng 南齋龔鈇校正). This

statement was reprinted in Zhang Zongxiang’s 1927 edition, but I have not been able to identify this

Gong Fu 龔鈇. Could he be the person who created the γ recension? Further research will be needed on

this issue.

Be that as it may, both internal and external evidence shows that the Shuofu must have been

expanded from the 366 titles of the Mao ms to the 725 titles of the mid-Ming mss not in one stage, but in

two stages. Or to put it differently, in between the α recension and the γ recension was a now-lost β

recension. Citations from Tao’s contemporaries confirm that Tao Zongyi did indeed compile a 100-juàn

Shuofu, although its 100 juàn did not include as many works as the sixteenth century one. When Yang

Weizhen praised the first version of the Shuofu as having “more than a thousand authors,” he was

engaging in literary hyperbole, since the first version had only 366. But within Tao’s lifetime, his Shuofu

38 Watanabe, “Setsu bu kō,” p. 230; King P’ei-yuan, pp. 5-6; Jao Tsong-yi, “Un inédit du Chouo-fou,” p. 94. On the

dating of the works, see Chang, Shuofu kao, pp. 366, 370, 386, and 388, and Xu Sanjian, “Jiguge cang Mingchao

liushijuan ben,” p. 115. The citation of these early Ming era works was embarrassing for Zhang Zongxiang who

originally claimed that his 100-juàn Shuofu published by the Commercial Press in 1927 was the work as Tao Zongyi

left it. He later acknowledged that the 1927 Shuofu edition actually included works of the Yongle era (1403-1424)

which must have been added after Tao’s death. His argument is that in these few cases, defective manuscripts must

have been supplemented by new sources. See Zhang Zongxiang, “Tieruyiguan suibi” 铁如意馆随笔, Zhonghua

wenshi luncong 中华文史论丛, 1984, no. 1, cited in Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” p. 115.

Detailed textual analysis of the SF text of the Xiao pin ji 效顰集 might be able to determine its date in relation to the

dated drafts of the work and hence its earliest date of incorporation.

18

had reached 100 juàn in size, and almost twice as many titles as before. Sun Zuo’s 1374 biography of Tao

describes his writings as follows:

Of late he has taken ever more to barring his gate and writing books. There are the Shuofu known

throughout the world in 100 juàn [or: Those known throughout the world are the Shuofu in 100

juàn], the Chuogeng lu (“Notes Upon Rest from the Plow”) in thirty juàn. . . .39

Similarly, Ye Sheng 葉盛 (1420-1474, jinshi degree 1445), in his Shuidong riji 水東日記, wrote a propos

Tao Zongyi:

Recently I have heard that the Shuofu in 100 juàn is still preserved in his family, without me

knowing which ones are the passages which Jiucheng has personally added or deleted. It is indeed

an incomplete work!40

These passages attest to a 100-juàn Shuofu personally created by Tao Zongyi, a text no longer extant, but

which I call the β recension. Of course, as Ye Sheng said, the Shuofu was essentially a collection of Tao

Zongyi’s private reading notes, and as such never had a completed and fixed form in his life. Ye Sheng’s

description of the manuscript describes a work which was still in progress up to his death, sometime not

long after 1401. Minor changes here and there in the β recension probably occurred frequently.

It seems, however, that after Tao died (shortly after 1401), his original β recension Shuofu was

then compressed into less than 70 juàn and 30 more juàn of works were added, creating the standard mid-

Ming γ recension Shuofu, with its 725 titles and 100 juàn. The only reference to this second

39Cited in Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 115-16; cf. the English translation in Mote, T’ao

Tsung-i, p. 31.

40Cited in Chang Bide, Shuofu kao, p. 19; Rao Zongyi, Xianggang daxue Feng Pingshan tushuguan cang shanben

shulu 香港大學馮平山圖書館藏善本書録 (Hong Kong: Lungmen Bookstore, 1970), p. 160; cf. the French

translation in Jao Tsong-yi, “Un inédit de Chouo fou,” p. 89.

19

reorganization comes from the fifteenth century writer Du Ang 都卬—a figure datable only from being

the father of the better known Du Mu 都穆 (1459-1525)—who described it from hearsay in his Sanyu

zhuibi 三餘贅筆: “The Shuofu was originally in 70 juàn; as for the latter 30 juàn, someone in

Songjiangfu took writings from the Baichuan xuehai 百川學海 and added them in.”41 Since many items

from the Baichuan xuehai are indeed found in the γ recension mss,42 it must be this “someone in

Songjiangfu” (who may also be one of Tao’s family who were keeping the manuscript according to Ye

Sheng or may be the mysterious Gong Fu mentioned in Fu1) who created the first exemplar of the γ

recension Shuofu that became common in the sixteenth century.

Chang Bide’s thorough analysis of the γ recension Shuofu confirms the essential accuracy of

what Du Ang heard.43 The γ recension44 has, as said, 725 separate titles. Of these, however, Chang found

72 to be also found in the Baichuan xuehai, and the vast majority of these Baichuan xuehai titles were

added in after juàn 67.45 This distribution indicates that up to juàn 67 of the γ recension is roughly the

same in contents as Tao’s original β recension, and that the β text had somewhere between 572 and 649

41 Cited in Rao, Xianggang daxue Feng Pingshan tushuguan cang shanben shulu, pp. 159-160. This observation was

repeated in the Siku quanshu zongmu tiyao 四庫全書總目提要 (Shanghai: Commercial Press, 1933), 123/2584; cf.

Pelliot, “Quelques remarques,” p. 175; Chang Bide, Shuofu kao, p. 13. The Baichuan xuehai was a thirteenth

century Song collection (cóngshū 叢書) containing integral texts; it was first printed in the Ming era.

42 Chang Bide, Shuofu kao, p. 15. 43 I have used Chang’s analysis on the placement of the Baichuan xuehai material in the 100-juàn Shuofu. But I

disagree with his interpretation of this data. I believe his mistake was to think that the editorial work of Yu Wenbo

(described in a preface of his which appears in the printed version of the Ming-Qing transition) had anything to do

with the production of the standard γ recension. In fact, no such Shuofu has any preface of Yu Wenbo’s. Chang Bide

relied on the fact that the T ms does mention Yu Wenbo, but my textual analysis proves beyond a doubt that the T

ms is actually a twentieth century copy done in awareness of the textual scholarship of the time. Yu Wenbo’s name

was likely added at that time. Thus the T ms is irrelevant to determining the nature of Yu Wenbo’s text. Reading Yu

Wenbo’s preface without preconceptions, it clearly applies to a different type of Shuofu than the standard γ

recension and he is presumably innocent of the charges of dishonesty directed at him by Chang. 44 Chang used the the T ms and Zhang Zongxiang’s 1927 Commercial Press edition. As far as the contents go these

are essentially identical also to the Zo, N, and Zg mss which I have examined personally. 45 See Chang Bide’s conclusions in Shuofu kao, pp. 13-22, esp. pp. 15-16. Juàn 1-67 contain 8 works out of 580

which can be traced to the Baichuan xuehai, while juàn 68-100 contain 64 works out of 145 which can be traced to

the Baichuan xuehai. Xu Sanjian, based on the research of Zhang Zongxiang, draws similar conclusions in his

“Jiguge cang Mingchao liushijuan ben,” p. 115.

20

titles.46 A division between Tao’s material and later material somewhere in the area between juàn 60 and

70 is confirmed by the fact that all the Ming works occur in juàn 84 or after, and that the last work whose

author attribution shows a Yuan loyalty is in juàn 57.47

The upshot of this discussion is that the SWQZL was copied into the Shuofu by Tao Zongyi as

part of the β recension. This occurred sometime between 1361 when the α recension of the Shuofu was

created and 1374, when Sun Zuo already knew of a much larger 100-juàn Shuofu. The subsequent re-

organization and additions made by the “someone from Songjiangfu” to create the γ recension did not

affect the SWQZL. Since the expanded version of the Shuofu was produced just around the Yuan-Ming

transition and included works on the rise of the Yuan it is tempting to speculate on how these reading

notes reflected Tao’s view of the dynastic transition. The second set of works anthologized in the Shuofu

contains a higher number of works related to border and overseas issues (see the sample in Table 1).48

Was Tao dealing with the fall of the Yuan by attempting to understand its legacy as a non-Han dynasty?

Or was he capitalizing on a spree of book sales as collectors sold off volumes on “barbarian” topics that

were no longer of interest in the new Ming dynasty? One also notices a relatively higher number of Yuan

authors in the second compilation (there were very few in the first compilation of the Shuofu).49 Here too

one may speculate about whether with the passing of the dynasty (whether imminent or very recent), Tao

was attempting to preserve some of its less well-known literary ventures. And finally, one may speculate

whether the presence of the Meng-Da beilu 蒙韃備録 and the SWQZL in the second compilation was due

46 It would have 572 titles if we assumed assumed that all Tao’s original material was moved to juàn 1-67, and the 8

out of 580 works in that part also found in the Baichuan xuehai were all interpolated. It would be 649 if we took the

725 of the standard 100-juàn Ming Shuofu and simply subtracted the four latest Ming works and all titles shared

with the Baichuan xuehai. The real figure is likely to be between the two, but closer to the lower figure. 47 See the list of Ming works above. The last work with a Yuan-loyal dating is in juàn 57/20b (Xue zhou cuo yu 雪舟

脞語, attributed to Song mo guo chu 宋末國初). The next works with a plain Yuan dating are Xu ji shan lu 續積善

録 (64/5a) and Jingxinglu 景行録 (64/6b). 48 Only three of these works, the Qidan guozhi 契丹國志 and Dajin guozhi 大金國志 in juàn 86, and the Liaodong

zhilue 遼東志略 in juàn 97 would likely have been added in during the posthumous reorganization that expanded

the number of titles from c. 600 to 725. 49 This may be most easily verified by scanning the author eras in the table of contents for Shuofu juàn 1-30 and 30-

67 in the catalogue of the Hn (Commercial Press) edition in the Tōhō bunka gakuin Kyōtō kenkyūsho kanseki

mokuroku, pp. 310-321 or of the T ms in Guoli zhongyang tushuguan shanben shumu, 2nd edition, vol. 4, pp. 1445-

84.

21

in part to Tao Zongyi’s realization that, despite his lukewarm loyalism, the Yuan had in fact fallen and its

taboos would never again be enforced. The Meng-Da beilu was a Song work of 1221 which described the

Mongols from the Song perspective, although sometimes quite positively, and used for them throughout

the word “Tatar” (Dada 韃靼), one forbidden under the Yuan as derogatory. On the other hand, the

SWQZL as an edited version of the Veritable Records of the dynasty was only allowed to be read by

official readers as long as the dynasty lasted. Both of these works would have been inappropriate for

public circulation during the Yuan itself. Despite the honoring of the anonymous author as a writer of the

Sovereign Yuan (Huang Yuan) inclusion of such works at the turn of the dynasty effectively marked the

Shuofu as a post-Yuan book.

Apart from the speculative questions about Tao as reader, bibliophile, and editor, the more

important questions about the Shuofu’s original copy of the SWQZL are whether it was a complete text,

and whether it was reliably copied. Song Lian said that whenever Tao anthologized a work he included

only the essential parts. Was the SWQZL abridged in that way and if so, how? Given the number of

almost incomprehensible passages left in the account of Chinggis Khan, it seems that he must have copied

the entire text on him that was available to him, despite not being able to understand it. (Was this a sign of

Yuan loyalism or of his interest in the bizarre and exotic?). For the text on Öködei, the situation is less

clear; given the extremely sketchy account of Öködei’s final years it is possible that Tao abridged

somewhat at that point. I think it is more likely, however, that Tao’s copy is complete of what he had and

that the abridgement of Öködei Qa’an’s reign is due to abridgment in his source. But Tao did presumably

abridge the work in the sense that the second juàn, titled the Qinzheng lu, and covering the reigns of from

Güyüg and Möngke to Qubilai Qa’an was eliminated without a trace, although its title appeared in the

SWQZL title.

As I will summarize below, and discuss at length in the chapter on the manuscripts, the SWQZL

text in the Shuofu underwent constant and cumulative corruption. Much of this process was random, but

much of it also driven by constant harmonization with the text of the YS. But in some cases, examination

22

of the text in comparison with YS and/or GH seems to indicate that even the archetype (that is, the most

ancient reconstructable text) of the SWQZL in the Shuofu is already significantly corrupted. Most of these

corruptions seem to be clearly just mistakes, while others seem to be cases of harmonization with the YS,

or attempts at improving the text.50 The original Mao ms was, as was mentioned, extremely sloppy and at

points almost unreadable. If the second batch of Shuofu materials were copied in the same way, then

much of the corruption in the SWQZL text would have entered in not due to later Ming-Qing copyists but

at the very beginning of the text’s transmission as a Shuofu work.

There is a further possibility that unfortunately cannot be confirmed or denied, due to the

remaining uncertainty of the date. The SWQZL was being copied into the SF sometime between 1361 and

1374. At the same time, the YS was being compiled from the Veritable Records by the historians of the

victorious Ming dynasty from 1369 to 1370.51 Chronologically, it is not impossible then that Tao Zongyi,

who was a good friend of the YS’s chief compiler, Song Lian, was actually aware of the YS as he was

copying the Veritable Record text into his Shuofu. This might account for some of the very early

instances of harmonization, for example, in which surnames are inserted for Jurchen and Kitan persons in

the SWQZL, or in which all the mss of the SWQZL share with the YS a corrupt text, for example, Hūlán-

Zhǎncè 忽蘭·盞側 (in YS 1.7 and SWQZL §15.3) for correct 忽蘭·虎惕.52 There are also two cases of

character variation found in the earliest mss, where harmonization with YS in the very earliest texts seems

rather likely. In these two cases, those of Sa’ari Steppe (薩里~撒曆) and Küchülüg Qa’an (屈出律~曲出

律), one type of transcription uses characters not found elsewhere for transcription in the SWQZL, but

which match that of the YS, while the other type of transcription uses common transcription characters.53

50 See the “Text Note on §40,” for example, on what I argue are old harmonizations. Most of the other notes, and

places in the critical apparatus where I cite the VR (Veritable Record) reading denote cases of corruption. 51 YS, appendices, “Jin Yuanshi biao” 進元史表, pp. 4673-74, and “Song Lian mulu houji” 宋濂目録後記, pp.

4677-78; cf. the “Introduction,” to Cleaves’s translation of the SHM, pp. xlv-l. 52 Of course in this latter case, the other possibility that the corruption occurred early in the Veritable Record and

was then handed on independently to both the YS and the SWQZL also cannot be ruled out 53 Sa’ari is found in the SWQZL in §§3.1, 14.4, and 16.1, each of which has a parallel in YS 1/3, 1/6, and 1/7. The

YS has 薩 throughout, the SWQZL has 薩 in §§3.1 and 16.1 but not in §14.4. Küchülüg is found in the SWQZL in

§§33.2, 36.2, and 47; the first two have parallels in YS 1/13 and 1/14 (cf. YSRMSY p. 458). YS 1/13 and 1/14 use

23

For these reasons, I tend to think that Tao Zongyi himself compared Chagha’an’s text to that of the YS as

he was copying the SWQZL into his SF. This would mean that he added the SWQZL to his SF only after

1370.

The Γ Recension and the Mid-Ming Book Trade Beginning in the Chenghua 成化 era (1464-1487), book production in the Ming dynasty entered a

sustained rise that would continue through the end of the dynasty and into the Qing. This rise gathered

momentum in the Zhengde 正德 period (1505-1521) and by the Jiajing 嘉靖 era (1521-1566) printing and

book production had reached levels orders of magnitude higher than those prevailing during the early

Ming.54 Although not printed until the very end of the dynasty, the Shuofu participated in this boom in

manuscript form. Except for the Mao ms and the fragmentary Fu1 ms, all other extant mss of the Shuofu

date to the Hongzhi 弘治 era (1487-1505) or later. Beginning in the Hongzhi era, editors also began

experimenting with new ways of improving the text of the Shuofu and repacking its structure, creating a

confusion of manuscript editions and texts that would continue until the present.

One of the most common ways of “improving” a Shuofu text was to find other exemplars of the

text being copied and borrow “good readings” from them. These other exemplars might be contained

within other Shuofu texts, but might just as well be independent of the Shuofu tradition altogether. Thus,

the SWQZL text as embedded in the Shuofu was often compared with the parallel text in Yuan shi, juan 1

and 2, and harmonized with it where it differed. Sometime before 1505 an anonymous editor of the text

did this in a massive way, albeit still within the context of a standard γ recension text. This editor was

working with something very close to the extant Zhang or Sun manuscripts of the Shuofu; indeed the

Zhang ms may be a draft made in the process of producing his text. From the exemplar he used, the editor

屈 in both cases; in the most primitive mss of the SWQZL, this is used only in §33.2 and elsewhere 曲 is used. More

derived mss harmonize usage with the YS change most or all of the instances of 曲 to 屈. The first instance of 屈

may well be a result of harmonization also. 54 Lucille Chia, “Mashaben: Commercial Publishing in Jianyang from the Song to the Ming,” in The Song-Yuan-

Ming Transition in Chinese History, ed. Paul Jakov Smith and Richard von Glahn (Cambridge: Harvard University

Press, 2003), pp. 303-06.

24

inherited several parablepses55 as well as a strange corruption that altered over half of the instances of dū

都 (commonly used in transcription to transcribe Mongolian -du or -tu) into xiāng 相. This changed, for

example, bádū 拔都, the standard transcription of Mongolian ba’atur “knight, hero” to the

incomprehensible báxiāng 拔相.

But building on a foundation much like the extant Zhang ms, the editor built a substantially

different edifice through harmonization with the YS. In addition to eliminating the honorific spacing for

Yuan imperial titles, the new editor noticed that an entry for the year yi/hai 乙亥 (1215) was missing from

the account. Thus he took what the Yuan shi had under that year and directly interpolated this 121-

character passage into the text at the end of §41. He did not notice, or perhaps did not care, that the events

described were mostly found elsewhere in the SWQZL, albeit in different versions. 56 He also began to

change the transcriptions to make them more like those of the YS, altering the transcription of Ong Qa’an

from 王·可汗 to 汪罕·可汗, Muqali from 木花里 to 木華黎, Altan from 按攤 to 按弹, Ïdu-Qut from 奕

都·護 to 亦都·護, and much more. The author made hundreds of other minor such changes through the

text, sometimes just making errors, but many times making difficult readings easy by harmonizing them

with the YS. The editor also made an idiosyncratic replacement in about half of its appearances as a

transcription character, of zhēn 真 with zhēn 貞; fortunately in this case the pronunciation was not

changed.

Whether because it made the SWQZL text more “readable” or for some other reason unrelated to

the editing, the SF manuscript with the resulting text in which these changes occurred—what I call the Hr

(for “harmonized”) exemplar—was quite successful, prolifically generating a large number of daughter

55 “Oversights” where the eye jumps from one character to the same character further down the text, thus eliminating

a whole chunk of text. These are particularly common in copying difficult texts like the SWQZL. 56 This inconsistency, while typical in reality of the attempts made to harmonize the SWQZL text as embedded in the

SF with the YS, confused William Hung into thinking that these sorts of changes, such as the interpolation after §41

must have been original to the SWQZL. See Hung, “Transmission,” p. 480 n. 116.

25

mss.57 In fact only three extant mss of the SF (Zo, N, and Zg) show no influence from this Hr exemplar.

The Fu3 ms, dated to 1505, shows the results of this editorial change, along with some additional

corruption, so this editing certainly predated that year.58

Perhaps due to the greater currency of manuscripts, one can also note a tendency for more

manuscripts to be produced by comparing two texts of the Shuofu. An editor, having made a copy of one

manuscript would then compare this copy to another, noting the different readings and substituting them

for those in his base text, where it seemed appropriate. At least in the SWQZL, this is generally done on an

eclectic basis, frequently prefer wrong readings in one paragraph and correct readings in the next. In a

few cases, however, such as the Uang ms, where a harmonized exemplar of the SWQZL text was then

collated with a text fairly similar to that of the Niu Shuofu, the result was a substantially improved text,

although still well short of what could be achieved with full access to a wider range of mss and the non-

Chinese parallels. Many of these later mss also attempted to reform the vulgar characters characteristic of

the older γ recension texts. Presence of older character forms may thus be a mark in Shuofu mss of a

relatively earlier mid-Ming date.

The Δ Recension Other editors directed their efforts towards reorganizing the Shuofu as a whole, either expanding the

number of texts or weeding out the redundancies or works found elsewhere. One of the best known of

these efforts to rework the γ recension was that of Yu Wenbo 郁文博. As he told the story, Yu Wenbo (b.

1418, jinshi degree 1454) acquired a 100-juàn (presumably a γ recension one) text in the Shanghai area

57 Ming mss deriving more or less directly from this episode of editing, without passing through any further major

changes include the Fu3 and She mss. Also the twentieth century Fu-Metropolitan and Taipei ms were copied from

the Fu3 (or perhaps She ms for the Taipei ms). I call these works the Fu-She family. The Pan text of the SWQZL also

appears to been copied two exemplars, one deriving from this Hr exemplar. See the descriptions of these mss. 58 In a previous article, I speculated that this editing may have been part of Yu Wenbo’s 郁文博 reorganization of

the Shuofu in Hongzhi 弘治 9 (1496). See Christopher P. Atwood, trans. Mukai Masaki 向正樹, “陶宗儀『説郛』

と『聖武親征録』『蒙韃備録』のテキスト伝承,” Tōyōshien 東洋史苑 77 (March, 2011), pp. 127-50.This was

in part due to reliance on Chang Bide’s linking of this editing with the Taipei ms, which I had not yet analyzed. As I

will show below, however, this Taipei ms is not a SF of the Yu Wenbo family, and neither are the other ones

showing these editorial changes. Thus there is no evidence linking this editing episode with Yu Wenbo.

26

near Tao Zongyi’s home around 1481. Examining the text, he noticed that it had numerous errors and

repetitions. As he lent his copy out to be copied by various officials in the area, negligent scribes allowed

further errors to creep in. Eventually, after retiring, he had a clean new copy made, at which point he also

decided to eliminate the 63 works in the Shuofu which were duplicated in the newly published Baichuan

xuehai. The remaining material he re-organized into 100 juàn, with a preface dated to the waxing third

moon of Hongzhi 9 (March, 1496).59

Assuming that Yu’s original text was a γ recension one, the conclusion from his preface would be

that he created a roughly 662-title, 100-juàn Shuofu. Since he mentions suffering loss in the process of

copying, the total number of titles was probably less than 662. In any case, it could not be identical to any

extant γ recension, with its 725 titles. Except for a concatenation of unfortunate historical accidents, this

conclusion would have been apparent to scholars long ago. The extant mss of the γ recension have no

visible connection whatsoever with Yu Wenbo’s recension. They have only Yang Weizhen’s preface, and

no preface by Yu Wenbo,60 their 725 titles, and 100 juàn show no trace of the reorganization discussed by

Yu Wenbo in his preface, and finally the extant and dated 725-title mss, particularly the Zhao ms of 1500

and Fu3 ms of 1505, show too much variation in their texts (at least of the SWQZL) to be plausibly

derived from a common ancestor dating as late as 1496. Add in the Fu1 ms associated with the

Congshutang library of Wu Kuan who died in 1505 and it is undisputable that the γ recension Shuofu

predates the year 1496, when Yu Wenbo says he created his new recension. All these considerations

should have made it clear from the beginning that the 725-title Shuofu had nothing to do with Yu Wenbo.

But since Yu Wenbo’s preface was attached to the block-printed Shuofu editions produced during

the Ming-Qing transition era, it was unfortunately assumed that his version must have been “the” textus

59Yu Wenbo’s preface is preserved in printed editions of the Shuofu; see Shuofu yibaiershi juan 説郛一百二十㢧,

pp. 1-3 in Shuofu sanzhong. It is also reprinted in Chang, Shuofu kao, pp. 13-14. Pelliot gave a French paraphrase in

“Quelques remarques,” pp. 170-174.

60 I have examined (directly or in photocopies) the prefaces and contents of Zhao, Niu, Zhang, and Taipei mss.

Zhang Zongxiang’s 1927 edition, based on standard 100-juàn Ming mss also lacks it (2nd preface “Shuofu xu” 説郛

序, pp. 1a-1b).

27

receptus of the Ming dynasty. It was thus also assumed that any attempt to reconstruct a version of the

mid-Ming Shuofu would be reproducing Yu Wenbo’s version. This assumption, implicit in the writings

of scholars like Pelliot and Jing Peiyuan, was made explicit by Chang Bide. Working in Taiwan, he had

access only to the 1927 Commercial Press edition of the SF and to a single manuscript kept in the Central

National Library (Guoli zhongyang tushuguan 國立中央圖書館) in Taibei (no. 000525628). As seen by

its published table of contents, this ms is a standard γ recension and it has only one preface, that of Yang

Weizhen.61 In the first page of juàn 1, however, it has the note: “Revised by Du 都 [sic, for Yu 郁]

Wenbo of Houxue 後學, Shanghai 上海.”62 From this note, Chang Bide drew the far-reaching conclusion

that the common γ recension mss were actually all descendants of Yu Wenbo’s ms of 1496.63 Needless to

say, this meant that Yu Wenbo’s preface suddenly seemed very misleading, not to say dishonest, since the

works he said he had eliminated were all included in his supposed manuscript!

The real origin of this note on Yu Wenbo has only emerged from my detailed examination of the

Taipei ms’s text of the SWQZL. As is I will demonstrate at length in my critical edition of the SWQZL,

this Taipei ms is not a Ming-dynasty manuscript, but a forgery produced after 1926 on the basis of

hybridizing an authentic Ming exemplar64 with Wang Guowei’s 1926 scholarly edition. This result for the

SWQZL text is absolutely indisputable. Given that fact, and the absence of a Yu Wenbo preface, it seems

unavoidable that at least the first juàn as well of the Taipei ms is a modern copy in which a reference to

Yu Wenbo was interpolated. And given the crudity with which the editor of T ms’s SWQZL text tried to

“improve” his copy (based on the Ming-era Fu3 ms, at least for the SWQZL) by following Wang

61 Guoli zhongyang tushuguan shanben shumu 國立中央圖書館善本書目, 2nd edition, vol. 4, pp. 1445-84. For a

photocopy of the preface and table of contents, I am indebted to Hsiao-ming Yu, Director of the Special Collection

in the Central National Library (Taipei), with the kind assistance of Indiana University East Asian librarian, Wen-

ling Liu. 62 I have not found a reference to Houxue 後學 as a village or as the name of Yu Wenbo’s residence, but there is a

Houxuecun 後學村 village in Qufu 曲阜 in Shandong 山東 province. 63 Chang, Shoufu kao, pp. 14-15 and Pl. 1. Jao Tsung-i followed Chang’s conclusion: see his “Un inédit du Chouo-

fou: Le Manuscript de Chen Han de la période kia-tsing (1522-1566),” in Mélange de sinologie offerts à monsieur

Paul Demiéville (1966), vol. 1, pp. 92-93. 64 This authentic Ming exemplar was the Fu3 ms for the SWQZL. But since Taipei ms has Yang Weizhen’s preface,

and the Fu1 ms does not, the forger must have used some other γ recension ms, not included Fu’s set of four mss

(Fu1-Fu4) as one of his base texts.

28

Guowei’s readings throughout the first half of the text, he cannot have had any scruples either about

adding a reference to Yu Wenbo, based, of course, on the well-known scholarly consensus of the 1920s

about the Ming Shuofu.65

As a result, it seems that the only certainly extant witness to Yu Wenbo’s recension is the Ming-

Qing printing of the Shuofu (to be discussed below), whose printers must have had access to his

manuscript since they included his preface. Unfortunately, the blockprinted edition does not include the

text of the SWQZL. However, it does include that of the Meng-Da beilu, and my preliminary examination

of that text indicates that the block-printed text of that work is independent of the γ recension texts I have

compared it with (Zhao, Niu, and Zhang mss), and preserves many older readings lost in other such γ

recension mss. This suggests that its text, like that of the blockprint’s preface, may derive from an

independent Yu Wenbo version of the SF. In that case, Yu Wenbo’s original copy of the Shuofu, which he

acquired in 1481 and made the base text for his 1496 manuscript, would be also independent of other

extant γ recension ms. Since there is no SF ms definitely known to be derived from Yu Wenbo’s, and

containing the MDBL, this suggestion of mine must remain speculative until stemmatic analysis of a

wider variety of texts in undertaken.

There is, however, one manuscript, from the Hūnan Printing House (Hunan shushe 滹南書舍),

which I think is likely to be also a copy of the Yu Wenbo Shuofu. This manuscript, which is the only

extant exemplar of the δ recension, has 100 juàn (of which only 55 survive), but the contents of these 100

juàn match only those of the first 60 or so juàn of the standard γ recension (see Table 2). The ms as far as

known generally follows the order of the γ recension, with a few exceptions in its juan 71-72. It is a

working copy with numerous proof-readers marks (〇 and 丶), as well as notes in the top margins and

65 The Taipei manuscript also eliminates in juàn 97 the two obviously Ming era works (the also Ming-era works in

juàn 84 and 87 are present); see Guoli zhongyang tushuguan shanben shumu, 2nd edition, vol. 4, pp. 1482-83, cf. p.

1480. Given that the manuscript is, even on its own claims, a middle Ming manuscript, I fear that this too is not an

indication of its earliness, but rather another crude attempt to make the manuscript seem more old and genuine than

it really is.

29

corrections between the lines.66 Fu Zengxiang also discussed it briefly, praising its good readings.67 Since

it includes three Yongle and later works--Quan shan lu 勸善録, Xiao pin ji 效顰集 (Hu’s juàn 80=γ

recension’s juàn 97), and Qianpu 錢譜 (Hu’s juan 70= γ recension’s juàn 84)—it cannot be earlier than

the mid-fifteenth century. Unfortunately, the first juàn, which would contain the prefaces to confirm my

proposed identity with Yu Wenbo’s recension, is missing. Likewise the remaining 55 juàn do not seem to

contain either the SWQZL or the MDBL, so I cannot currently say anything about its stemmatic position.

But the overall organization and date seems similar to what is described in Yu Wenbo’s preface.

Moreover, of all the works in the δ recension I have been able to identify, none are found in the Baichuan

xuehai which is exactly to be expected if the δ recension is in fact the Yu Wenbo recension. The ultimate

proof of this identity, however, can only come from close comparison of Hu ms texts with those of the

blockprinted one. Since the blockprinted edition did have access to the Yu Wenbo recension and cited its

preface, there should be some works at least, where the blockprinted text is more or less identical to the

Hu ms text, but rather different from, and perhaps superior to, the γ recension mss. If on the other hand

after comprehensive textual examination, there are no such cases where the blockprint texts align with the

Hu ms, then this identity I am tentatively advancing here would have to be rejected.

The Ε Recension Another Shuofu with a reorganized format is extant in two manuscripts, one kept in Hong Kong and one

in Beijing. The Hong Kong ms, designated the Shen ms, is a 69 juàn manuscript first described in 1970

by Jao Tsung-i which was copied for Shen Han (jinshi degree, 1535), and now held in the University of

Hong Kong’s Fung Ping Shan Library.68 The one in Beijing has catalogue no. A01507 in the National

66 The presence of both original text and corrected text opens intriguing possibilities. Was the extant ms being

collated with some other ms? If so, can that ms’s text be indentified? The fact that it was written on stationary of a

publishing house might also suggest that an otherwise unknown blockprint edition of the Shuofu was at some point

contemplated. Much remains to be researched with this manuscript. 67 Mo Youzhi, Cangyuan dingbu Lüting zhijian zhuanben shumu, p. 752 (second Shuofu listed). 68 First described in Jao Tsong-yi 1966; Rao Zongyi [Jao Tsung-i], 1970, and at somewhat greater length in his 1982

“‘Shuofu’ xin kao—Ming Jiajing Wujiang Shen Han chaoben ‘Shuofu’ jilue,” reprinted in Rao Zongyi 1993. This

manuscript is also described in Fu Zengxiang’s supplements to Mo Youzhi, Cangyuan dingbu Lüting zhijian

zhuanben shumu, p. 752, as the last of the Shuofu listed.

30

Library, and is called the Shi ms.69 Together these two mss, whose texts of the SWQZL are extremely

close to each other, form the ε recension of the Shuofu.

The two manuscripts of this ε recension lack both a table of contents or numbering of the juàn.

This absence of a table of contents has made it very inconvenient for scholars to give a full description of

the contents of such a manuscript; to date no one has. Jao Tsung-i, however, did describe the contents of

the some of the volumes of the Shen ms, and I have given the organization of that part of the Shi ms

before and after the SWQZL text. The results of both show the organization to be completely different

from that of the standard γ recension (see Tables 3 and 4).70 Another striking feature, which does not

show up in a comparison of the contents, is the identical mise en page of the two mss, in which each

character is placed in exactly the same place in the column—where one or the other ms (usually Shen,

which was more carelessly copied) omits a character, another character is duplicated at the column foot to

keep the same alignment. Where Jao describes the arrangement of the sub-titles in collective chapters

such as Zhuzi suishi 諸子隨識 (“Random Opinions from the Great Thinkers”) or Zhuzhuan zhai xuan 諸

傳摘玄 (“Notes on the Occult from Biographies”), the arrangement in the Shen Han ms is even further

from that of the Mao ms than is that of the standard γ recension manuscripts.71 The creator of this ε

recension was evidently moving toward a content-based topical organization.72 Moreover, while the 69

juàn is close in number to the 70 which Du Ang said was the number of juàn in the original Shuofu, the

actual number of titles is much closer to that of the γ recension than to the Mao ms.

69 Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji shanben shumu: Zi bu 北京圖書館古籍善本書目·子

部, p. 1695. 70 There is some inconsistency in the organization between the two mss. In the Shen ms, the SWQZL is immediately

followed by the Bei yuan lu 北轅錄, while in the Shi ms, the Bei yuan lu 北轅錄 is in the 10th fascicule, preceding

the 11th fascicule which contains the SWQZL. But given the fact that the Shi ms is extremely fragmentary, while the

Shen ms is virtually complete, I guess that in this case, the Shi ordering is a result of rebinding of a broken up ms. 71 See Rao 1993, p. 661. 72 In the Mao ms, for example, Tao Zongyi began with 14 juàn of cóngshū or collectanea—smaller collections now

to be included in a larger one. In the later 69-70-juàn MS, he seems to have taken the material on the Confucian

canons (jīng 經), previously in juàn 34-35 and given them pride of place in juàn 1 and 2. The ε recension, on the

other hand, put those works concluding in the character jīng 經, many of which were recently written canons of taste

(wine, horse-riding, etc.), not classics, in juàn 1.

31

Both the Shen and Shi mss include the SWQZL in the extant portions.73 The text in these two mss,

which I designate the ε recension, is also very distinctive. Compared to other SWQZL it has three major

features: 1) its base text is a close descendant of the Hr exemplar, the massively harmonized γ recension

exemplar that was also used for the Ming-era Fu3, Uang, and She ms. 2) This Hr exemplar text was,

however, then collated with a now-lost primitive exemplar of the Shengwu qinzheng lu that was

circulating independently of the Shuofu.74 3) Finally, it was harmonized once again to a truly exceptional

degree with the YS, with 23 larger or smaller interpolations from the YS text not shared with any other

manuscripts, as well as many smaller text changes. As a result the ε recension texts present one of the

most distinctive appearances of any SWQZL text. The occasional readings from the primitive, non-Shuofu

exemplar are extremely valuable, but often hidden by these massive harmonizations and idiosyncratic

readings, particularly in the Shen ms which in the transcriptions of Mongolian names, made several

further choices, such as altering zhá 札 to qǐ 杞 or miè 滅 and miè 蔑 to fá 茷.

It is unknown when and by whom the first ms of the ε recension was created. The Shen Han ms is

dated to the Jiajing era (1521-1566), but it must be at least once copy away from the common ancestor of

the family. However, the ε recension texts are not particularly primitive. As I mentioned, that of the

SWQZL is clearly based on a copy of the Hr exemplar, which is itself a late-fifteenth, early sixteenth

century branch off the common family tree of γ recension. Likewise with regard to the Meng-Da beilu

text my preliminary collation shows that the Shen ms text has the most numerous shared innovations

(synapomorphies) compared with those from other γ recension texts (mostly closely with Zhang and more

distantly with Zhao and Niu). The textual evidence is thus clear: the ε recension’s organization is not, as

Jao Tsung-i thought, evidence of its primitivity but rather a result of taking a standard γ recension

manuscript and reorganizing it in a way that would seem closer to Tao Zongyi’s original intention.

73 Contents of the two mss are given in the tables. Unfortunately none of the contents of the very fragmentary Shi ms,

of which only a fourth or a fifth of the original SF overlap with the very cursory description of the contents of Shen

given by Jao. But both are clearly very different from the γ recension. 74 This primitive exemplar was allied to another non-Shuofu ms also used by the ancestor of the Lu 陸 ms (Seikadō

Bunko 靜嘉堂文庫 Library, Tokyo) and Zheng 鄭 ms (National Library of China), both of the SWQZL alone.

32

Presumably the editor was working in a reading market already familiar with Ye Sheng and Du Ang’s

doubts about the original form of the text, and perhaps Yu Wenbo’s reorganized text as well. The

comments of these well-known bibliophiles primed the book market for a 70-juàn Shuofu. To those

rendered suspicious of the Shuofu texts, the ε recension could seem like something much superior to the

“parasitic additions and random overturning of the order” supposedly characteristic of other Shuofus.75 In

reality it was the SS text creator who was guilty of such bibliographic offenses.

The Ζ recension Given the growing interest in the Shuofu, it was only a matter of time before someone would think of

block-printing the work.76 Sometime before 1621, the Wanweishan Tang 宛委山堂, a Hangzhou printing

house, tried to produce the first printed edition of the Shuofu.77 As the core of the new work, the printers

must have had access to a rare manuscript in the Yu Wenbo tradition (thus, as I hypothesize, of the δ

recension) since they included his preface, but that manuscript was perhaps incomplete, and in any case

did not include all the 725 titles readers would have expected. So like other new versions of the Shuofu,

75 See the comments of the Shen Han ms’s owner Lu Qiao 陸樵 (fl. ji/chou 己丑, probably 1589) cited in Jao, “Un

inédit de Chouo fou,” p. 91; Chinese text cited in Rao 1993: 657. Was he basing his comments only Du Ang’s

doubts about the standard 100-juàn Shuofu? Or was he also aware of the Yu Wenbo recension and its preface? 76 The most convenient access to the blockprinted edition is that in Shuofu sanzhong “Three Kinds of Shuofu.” The

“three kinds of Shuofu” reprinted are Zhang Zongxiang’s 1927 printed version of the 100-juàn Shuofu, the 120-juàn

blockprint Shuofu printed of the Ming-Qing transition, and the 46-juàn Shuofuxu 説郛續 or “Shuofu Sequel” that

was included along with the Ming-Qing transition printing. But it is important to note that the edition printed in

Shuofu sanzhong is not a facsimile of any actually existing printing, but a composite reprint, mixing copies of a late-

Ming printing with the early Qing prefaces. The actual contents reprinted and the arrangement of the text is that of

the first, Chongzhen 崇禎 era, printing, identical to that of the copy preserved in the Institute of Oriental Culture in

Kyoto. Thus it includes the Inner Asian works deleted in later printings. But the printing also includes the Shunzhi

順治 3 (A.D. 1646) prefaces by Li Jiqi 李際期 and Wang Yingchang 王應昌, which were added only after those

Inner Asian works were deleted. It is also worth noting that the catalogue entry for the Kyoto copy found in the

Tōhō bunka gakuin Kyōtō kenkyūsho kanseki mokuroku 東洋文化學院京都研究所漢籍目録, pp. 324ff also

includes the name of Tao Ting 陶珽, the Li Jiqi preface and the Shunzhi 3 date even though none of these things are

found anywhere in the edition, and the copy is unquestionably a Chongzhen era one, as was long ago determined by

M.K. Watanabe. 77For what follows I have relied entirely on Chang Bide’s masterful detective work; see his Shuofu kao, pp. 25-35.

33

the printers must have used multiple manuscripts to produce the text. The result was a final version, the ζ

recension, whose precise manuscript affiliations are still unclear.

Already the manuscripts in the Shuofu tradition were sufficiently corrupt that they sought to use

printed versions of the items taken from other anthologies wherever possible. The hunt for new works

was incessant, and the new volume was expanded to 120 juàn, with an additional “continuation” (xù 續)

in 47 juàn. In Yang Weizhen’s preface the character yī 一 had been changed to èr 二, whether

deliberately or by corruption, and now Tao Zongyi was said to have based the Shuofu on his reading of

2000, not 1000, authors! To fill up this mythical number, the printers scoured anthologies to incorporate

new works, chopped up large works into separate one, duplicated works under alternate names, and even

listed unavailable works they thought Tao Zongyi might have included in the table of contents with the

notation “missing.” The new edition was given a topical organization, with most travelogues to the north

or east being grouped in juàn 55 and 56, and descriptions of remote provinces and Southeast Asian

regions in juàn 62. The effort to increase the number of works to 2000 resulted in the inclusion of a

number of interesting works on foreign peoples not included in Tao’s original anthology (see Table 1).78

Just before the work was to be published, however, the 1621 fire destroyed much of Hangzhou.79

All or most of the wooden printing blocks survived, but the publishing house could no longer fund the

printing and the wooden blocks were sold off to other printing houses in Hangzhou, where with a little

altering they were used as part of the printing for six different other anthologies. Finally in the Chongzhen

崇禎 era (1628-1644) of the late Ming, the Wanweishan Tang 宛委山堂 press recovered the blocks and

made two separate printings of a Shuofu edition in 120 juàn, with a 46-juàn sequel anthologizing Ming

78See for example, Liu Yu’s Xishiji on the Il-Khanate; Shi Maoliang’s Birong yehua on the Jurchen Jin; Jin, Hu

Jiao’s Xianlu ji on the Kitan Liao; Song Yande’s Gaochang jixing on Uyghuristan; and Fang Feng’s Yisu kao and

Xu Jing’s Shi Gaoli lu on Korea.

79Chang believes that there is a literary reference to a pre-1621 test printing, but that no actual copy of this printing

has survived to the present (see Shuofu kao, p. 27-28).

34

works. The total number of titles was around 1,360.80 The prefaces were those of Yu Wenbo and Yang

Weizhen, together with a notice from the publisher on “reading the Shuofu” (dú Shuōfú 讀説郛). Yet

whether because they were missing in all the manuscripts, or because their blocks were never recovered, a

small number of works from the γ recension did not make it into any of the block-printings; among them

was the Shengwu qizhenglu.81

The turmoil of the Ming’s fall and the campaigns of the new Manchu Qing dynasty (1636-1912)

to conquer the Ming territory prevented this first printing from gaining currency. Under the new dynasty,

moreover, works such as had earlier been included in the Shuofu that reflected Song attitudes to the

Kitans, Jurchen, and Mongol regimes were problematic. Printing works cursing rulers from Manchuria as

running a “cowards’ court” (Luting 虜廷)—and worse—was not something a prudent publisher would

risk. So when the same press reprinted the work in 1646 and the Shuofu finally became a widely available

work, it retained neither the SWQZL nor the Meng-Da beilu. Of the 14 works on the Kitan Liao, Jurchen

Jin, and Mongol Yuan dynasties found in the first Chongzhen printing, only three or at most four found

80Chang Bide identifies the printed edition purchased in 1943 by the Centre Franco-Chinois d’Études Sinologiques

in Beiping and discussed by King P’ei-yuan, Études comparative, pp. 6-9, as the very first known printing. A

complete table of contents (based on cat. no. 4104-87-3560) is published in Seikadō bunko kanseki bunrui mokuroku

靜嘉堂文庫漢籍分類目録, pp. 966-990. The Ming print kept in Kyoto’s Institute of Oriental Culture and published

by Watanabe Kozo in his “Setsu-bu kô” in 1938 he identifies as a later, slightly expanded, printing made by 1643 at

the latest. Its contents have been published in the Tōhō bunka gakuin Kyōtō kenkyūsho kanseki mokuroku, pp. 324-

47. The first printing had 120 juàn and 1360 titles (of which 124 were labeled “missing”) together with 44 juàn of

continuation (containing 544 titles of which 6 were labeled “missing”), while the second had 1364 titles (of which

113 were labeled “missing”) together with 46 juàn of continuation (containing 542 works of which 8 were labeled

“missing”), of continuation. See Chang, Shuofu kao, pp. 30-31.

81Among the works listed above as dealing with topics outside China proper, only the SWQZL, the Shi Liao

lu (“Record of an embassy to the Liao”) and the Qingtang lu on Kökenuur of all the works found in the 100-juàn

version appear to be missing from the early blockprint version. Chang Bide counted 206 works found in the original

725 titles of the 100-juàn Shuofu which are not found in block-printed version (p. 30). Wang Zhouyao, writing in

1917, compared the 1361 Mao ms with the block-printed edition and found over 860 works in the block-printing that

were not in the Mao ms and somewhere over 100 works in the MS that were not in the printed work (cited in Xu

Sanjian, “Jiguge cang Mingchao liushijuan ben,” p. 113). If this is the case, then works on foreign topics show an

unusually high rate of retention in the Ming blockprint edition. It is quite possible, however, that lacking the

convenient index found in volume 10 of the Shuofu sanzhong edition which I used, that there are works found in

which they missed.

35

their way into later Qing-era printings.82 Edited by Li Jiqi 李際期 (jinshi 1640), some printings added

new prefaces by Li himself, as well as by Wang Yingchang 王應昌, while others stuck with the previous

printing’s assemblage of front matter.83 Li Jiqi’s mutilated edition was in turn made the basis for text of

the Shuofu in the Manchu Qing dynasty’s colossal imperial literary compendium, the Siku quanshu

(“Complete Library of the Four Treasuries”) of 1772-1794. Until 1927, it remained the standard text of

the Shuofu, despite its vast divergence from Tao Zongyi’s original anthology.84

The printing of the Shuofu slowed the production of further manuscripts of the work as the block-

print versions came to define what people meant “the Shuofu.” Manuscripts did survive and were

occasionally consulted. Thus in 1778, comparing a separate ms of the SWQZL, Zheng Jie had a friend

copy a text of the SWQZL which he found in a manuscript Shuofu he located in a collector’s near

Changmen 閶門 Gate in Suzhou. There is no further information from which one could identify the

particular recension of the Shuofu in question, but judging from the readings which Zheng recorded, the

text of the SWQZL is one not otherwise attested in any Shuofu text.85 This interesting manuscript is an

illustration both of how much is still unknown about the Shuofu textual tradition and of many fragmentary

manuscripts may still remain to be discovered and identified.

82 Based on my personal examination of four copies in the Toyo Bunko (cat. nos. V-5-A-11, 12, 13, and 14) and one

Qing printing in the Seikadō Bunko (cat. no. 8505-163-305-2). The contents of a typical such bowdlerized Shuofu

blockpring can be found in Seikadō bunko kanseki bunrui mokuroku, pp. 990-1014, based on cat no. 8505-163-305-

2. Cf. King P’ei-yuan, Études comparatives, p. 15. Note that in no case were the actual whole blocks re-carved; the

variation was simply one of using or not using particular blocks and in the table of contents carving out certain

offending titles and gluing in strips with the revised titles carved on to them. 83 Toyo Bunko, no. V-5-A-11 has the old front matter, while Toyo Bunko no. V-5-A-12 and Seikadō Bunko no.

8505-163-305-2 used the new version with the Li Jiqi and Wang Yingchang prefaces. Those with the new front

matter also inserted a reference on the first contents page to Tao Ting 陶珽 (from Yao’an 姚安 in Yunnan 雲南,

jinshi degree, 1610) as having re-organized the text. As Chang argues, however, the fact that the biographical

sources on him seem to know absolutely nothing of any such enterprise on his part cannot be explained away and

makes his involvement very uncertain (see Shuofu kao, pp. 22-25). 84 A composite text, including the new front matter of the Li Jiqi printing, but the full contents of the second

Chongzhen-era printing of the Shuofu and Shuofu xu (“Sequel to the Enclosure of Literature”) was reprinted in 1988

as the third to tenth volumes of the Shuofu sanzhong.

85 It is, however, very similar to the Wāng 汪 text of the SWQZL kept in the Nanjing Library.

36

The 1927 Commercial Press Edition

As I have already mentioned, the 1927 Commercial Press edition of the Shuofu was a milestone, in which

the γ recension, once dominant in the manuscript but since the end of the Ming dynasty eclipsed by the ζ

recension, returned into scholarly view. Edited by Zhang Zongxiang, the Commercial Press SF is

essentially identical to the Zhang mss in contents.86

Zhang Zongxiang gave virtually no information about how he edited the volume, except for a

brief listing of the mss in a colophon at the end of his printed edition. Dated to ren/xu 壬戌 (A.D. 1922),

the colophon lists six mss of the 100-juan SF, all incomplete, which he claims to have used. 87 The first,

from the Metropolitan Library (京師圖書館) covered up to juan 32, the next was the composite Fu ms,

currently held in the Shanghai Library, while the third, kept in the Hanfenlou, is what I follow Jia in

designating as the Zhang ms (from Zhang Yuanji, who first described it). The last ms, the Sun ms, he

implies he used only to make up the juan missing from the others. In other words, the implication is that

for each work in the SF, his edition is based on a single ms. Of these mss, the latter three are all extant,

but the first is no longer extant, to my knowledge.88

Jia Jingyan already noted, however, that there is something puzzling about Zhang Zongxiang’s

recension of the SWQZL. First he pointed that of all the mss Zhang listed, it is the Fu3 ms from Hongzhi

18 which contains the SWQZL, Jia continues:

86 This edition is widely available in the original 1927 printing and in a photographic reprint in the Shuofu sanzhong

edition of 1988. 87 See Shuofu, 1927 edition, ba 跋, pp. 1a-b; Shuofu sanzhong, p. 1358c-d. 88 Can the contents of this ms be reconstructed? To a certain extent I believe they can be. As I mention below, there

is a copy in the Zhejiang Library of the ms Zhang Zongxiang used while compiling his edition of the Shuofu. In this

ms his base text (with occasional editorial emendations) is written in black, while collated readings from other mss

are added in red ink. By comparison with the other mss which Zhang used and which are extant (Fu1-4, Zhang, and

Sun), one could presumably isolate those titles in the Shuofu whose texts clearly differ from any of those three. Such

texts would then be presumptively derived from Zhang’s first ms.

37

Yet this reprinted Shuofu text’s Qinzheng lu is not similar to the Fu text in Wang Guowei’s

commentary or as recorded by Pelliot, nor is it similar to the Zhang text. Where the characters are

inferior, it looks like it is in between them, so where in the world did it come from?89

Stemmatic analysis resolves Jia’s puzzlement, by showing that the 1927 recension was based not

on a single ms but on a collation of two mss, the Zhang ms and what Jia called the Metropolitan ms,

because it was written on stationary of the Metropolitan Library in Beijing.90 This second ms in turn is a

rather poor modern copy of Fu3, occasionally collated with the earliest scholarly edition of the SWQZL,

printed in 1894. For this reason, I prefer to designate it the Fu-Metropolitan ms. This stemmatic analysis

is confirmed by the draft manuscript prepared by Zhang for his printed edition. Now held in the Zhejiang

Library (no. 7437),91 this manuscript preserves Zhang’s base text, written in black, which for the SWQZL

corresponds largely to the Fu3 ms, but with some minor editorial emendations. In red ink, however,

Zhang added readings taken from the Zhang ms. This evidence demonstrates that the texts in the 1927

edition are not, as one might expect from Zhang’s description of his practice, each simply taken from one

particular manuscript. Rather they are as a rule hybrid texts eclectically merging two or more γ recension

mss. This hybrid nature was, at least in the case of the SWQZL, somewhat less visible, because the mss he

used for it, the Zhang and Fu3 mss were already quite close to each other.92 In other cases, where the mss

chosen are less obviously related, the hybrid nature of the 1927 edition readings might be more obvious.

89 Jia, I, zhuiyan, p. 4a. 90 This Fu-Metropolitan ms was kept in the National Library, where it was used by Jia, and included the SWQZL in

juan 55; it should not be confused with the one Zhang himself used, which was kept in the Metropolitan Library, but

did not go beyond juan 32. 91 A complete table of contents of this ms is given in Zhejiang Tushuguan guji bu 浙江圖書館古籍部, ed., Zhejiang

Tushuanguan guji shanben shumu 浙江圖書館古籍善本數目 (Hangzhou: Zhejiang Educational Press, 2002), pp.

670-80. 92 As I will demonstrate in my critical edition of the SWQZL, the two mss are very close because in the Ming

dynasty’s Hongzhi era when the Fu3 ms was being created, the editor already used the Zhang ms or something very

like it to collate the ms.

38

Conclusion

The conclusions presented here are only the beginning of the analysis of the texts of the works included

with the Shuofu. They are based on a detailed analysis of the SWQZL and a preliminary analysis of the

MDBL and the Yang Weizhen preface. As a result there are many questions still unanswered or for which

the answers are only tentative. These include the proposed identification of the δ recension with that of

Yu Wenbo, and the precise mss which were used to create the ζ recension.

Moreover, there is no guarantee that all the works within a given manuscript of the Shuofu have

the same stemmatic relationship to those in other Shuofus. My preliminary examination of the MDBL

indicates that a stemma based on its text would indeed match that of the SWQZL. However, it was not

uncommon for a particular work in a Shuofu ms to be copied not on the basis of a single exemplar, but of

two or more exemplars. Thus of the SWQZL texts found in the Shuofus I have examined, the Fu3, Uang,

Shi, and Shen mss all show evidence of having collated one base text against another manuscript. In other

words, scholars who produced these mss had access to more than one Shuofu text, and it is always

possible that they might have preferred one ms for one title with the Shuofu, and another ms for another.

Moreover, it is unlikely that Ming-era Shuofu mss were always complete. Someone wishing to make a

complete Shuofu text in the sixteenth century might well have been reduced to the same expedient as Fu

Zengxiang or Zhang Zongxiang in the twentieth century: cobbling together 100 juan with a wide variety

of more or less fragmentary Shuofu mss. A copy based on such fragmentary Shuofus would show

differing stemmas depending on the title chosen for analysis.

Thus only detailed analysis of each of the 725 or so titles contained within the Shuofu can

eventually give a complete picture of the development of this complex collection of texts. One further

benefit of such a broad-based study of each text is that it will allow the identification of the large number

of very fragmentary Shuofus currently kept in Chinese libraries, containing only five or ten juan, or even

fewer. It is not impossible that among them may be found fragmentary exemplars of extremely valuable

mss. Identification would have to proceed one by one, however, based on an understanding of the stemma

39

of each particular title as derived from the better preserved and better known exemplars I have discussed

in this article. Such a colossal task is obviously beyond the abilities of any one scholar, and can only be

the goal of a team effort, drawing scholars interested in the full range of topics covered by Tao Zongyi’s

eclectic interests. It is to be hoped that this small preliminary analysis will be helpful to anyone thinking

to undertake this great task.

40

Appendix: Mss of the Shuofu

Extant and Catalogued Shuofu Mss 1. Mao 毛：Currently held in the Linhai City Museum. Α recension. 60 juan, completely extant in 20

fascicules. The SWQZL is not included. A full table of contents of the manuscript has been published.93

One preface, by Yang Weizhen 楊維楨; contents organized by final character and radically different from

the 1927 Commercial Press edition. The ms is written on unlined paper with no “fishtail” or running

header. The text has 9 columns per page and 17 characters per standard column. There are corrections

both in black ink, probably by the original editor, as well as in red ink by subsequent owners.

History: The ms is generally believed to be a Ming-era one. The inconsistency in character forms,

particularly for the rare character zhóu 㢧, which Tao Zongyi specially chose to replace the more usual

juan 卷, would seem to indicate that it is at least several copies away from the original 1361 copy

prepared by Tao Zongyi. The earliest known owner is Mao Yi 毛扆 who made corrections in red ink and

left a colophon in juan 20 with his stamp “Yushan Maoyi shougao 虞山毛扆手校.” It was then acquired

by Ma Yutang 馬玉堂 (courtesy name Huzhai 笏斋, sobriquet Qiuyao 秋藥, degrees 1821 and 1845) and

in the Tongzhi era by Wang Yongni 王咏霓 (1839-1916, courtesy name Zichang 子裳, sobriquet Liutan

六潭, jinshi degree 1880), who deposited it in the Jiufeng Shuyuan 九峰書院 (later known as the

Huangyan Jiufeng Library 黃嚴九峰圖書館) in Taizhou 台州 (modern Linhai). His landsman Wang

Zhouyao (1855-1925, courtesy name Meibo 玫伯, sobriquet Mo’an 默庵) also made further proofreader’s

corrections in red ink and added a colophon following the preface.94

93 See Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 118-27. 94 See Xu Sanjian, “Jiguge cang Mingchao liushijuan ben,” pp. 112-15, and “Shanben miji lun ‘Shuofu’” in Zhou

Xiangchao and Xu Sanjian Lishi wenhua mingcheng Linhai, pp. 245-48. The later has photographs of several of the

stamps and colophons of the ms.

41

2. Zhao 趙: Currently held in the National Library of China (no. 3907). Γ recension. 61 juan

extant in 50 fascicules. The SWQZL forms juan 55 of text.95

One preface, by Yang Weizhen 楊

維楨; contents show only occasional minor differences from the 1927 Commercial Press edition.

The paper is lined in blue with two “fish tails” and blue header blanks on the page fold; however,

the spaces have neither numbers nor any running header.

History: At the end of juan 24, Jia Jingyan found the following note, “copied in Hongzhi year geng/shen”

弘治庚申依本錄, thus dating the copy to 1500. It contains stamps and inscriptions from a large number

of scholars: Mr. Zhao from Wu (吳郡趙氏), Ruan Yuan 阮元 (1764-1849), Weng Binsun 翁斌孫 (1860-

1922), and others.96

Jia Jingyan speculates that the Mr. Zhao from Wu might Zhao Huanguang 趙宦光

(Wanli 萬曆 era, 1572-1620) or his son Zhao Jun 趙均 (Chongzhen 崇禎 era, 1627-1644). Weng’s

collection was the immediate source before it was acquired by the National Library.

3. Niu 鈕: Currently held in the National Library of China (no. 2408). Γ recension. 97 juan extant,

grouped in 70 fascicules; the SWQZL forms juan 55 of text.97 A full table of contents of the manuscript

has been published.98 One preface, by Yang Weizhen 楊維楨; the contents show only occasional minor

95 Beijing tushuguan shanben bu 北京圖書館善本部, ed., Beijing tushuguan shanben shumu 北京圖書館善本書目

(Beijing: Zhonghua shuju, 1959), vol. 5, p. 37a; Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji

shanben shumu: Zi bu 北京圖書館古籍善本書目·子部 (Beijing: Catalogue and Bibliography Press, 1987-88), p.

1694; Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed., Zhongguo guji

shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989), 1.38a (p. 75 in the

continuous pagination added to the 1990 reprint); Weng Lianxi 翁連溪, ed. Zhongguo guji shanben zongmu 中國古

籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部, pp. 1927. 96 Other owners listed by Jia Jingyan include: Zhang Ruizhong 張睿鐘, Xu Tieyi 徐鐡彜, Zhao Yuanxiu 趙元修,

Wei Quji 衛去疾, and the private libraries Changshu shezhuang 常熟捨莊, Mr. Yang’s Shanqing Hall 楊氏善慶堂,

Zhou Jianqi 周鑒齊，Xiao Hanjian 削漢劍, and Wei Weidou 魏慰斗主人. 97 Beijing tushuguan shanben bu 北京圖書館善本部, ed., Beijing tushuguan shanben shumu 北京圖書館善本書目,

vol. 5, p. 36b; Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji shanben shumu: Zi bu 北京圖書館古籍

善本書目·子部, p. 1694. 98 See Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed., Zhongguo guji

shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989), 1.10a-24b (pp. 19-

48 in the continuous pagination added to the 1990 reprint), and Weng Lianxi 翁連溪, ed. Zhongguo guji shanben

zongmu 中國古籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部, pp.

1919-1923.

42

differences from the from the 1927 Commercial Press edition. The paper is lined in blue with no “fish

tails” but white header blanks on the paper fold; the spaces are not numbered but have the running header

Shixue lou 世學樓. This was the residence of the Ming collector Niu Shixi 鈕石溪 of Shaoxing 紹興,

who therefore was responsible for the copying of this SF ms.99

History: Jia Jingyan notes that this ms was held by He Zhao 何棹 and Chen Kui 陳揆 (courtesy names

Zizheng 子正 and Zhun 准; 1780-1825) of Changshu 常熟. The Qing-era catalogue Lüting zhijian

zhuanben shumu 郘亭知見傳本書目 also mentions this ms as being in the possession of Chen Kui. Jia

further notes that most of Chen Kui’s library derived from that of Qian Qianyi 錢謙益 (1582-1664), also

of Changshu, and that the catalogue of Qian’s private library mentions a SF in 100 juan which Jia

believes should be this ms. 100

Sun 孫: Currently held in Yuhailou 玉海樓 museum, Rui’an 瑞安 city, Zhejiang province. Γ recension.

52 juan extant, bound in eighteen fascicules; the SWQZL forms juan 55 of text. 101 The paper is lined in

99 Although he seems widely known, I have not yet been able to identify his era. 100 Jia Jingyan, Shengwu qinzheng lu jiaoben 聖武親征錄校本 (unpublished ms, 乙未), I, zhuiyan, 3b; Mo Youzhi

莫友芝, supplemented by Fu Zengxiang 傅增湘, ed. Fu Xinian 傅熹年, Cangyuan dingbu Lüting zhijian zhuanben

shumu 藏園订補郘亭知見傳本書目 (Beijing: Zhonghua shuju, 1993), vol. 2, 10B/751; Qian Zeng 錢曾, Shugutang

cangshu mu 述古堂藏書目 (rpt. Taipei: Yiwen, 1965), 2/17b. However, as noted by Jao Tsung-yi, the N ms in the

National Library is bound in 70 fascicules while that recorded in the Shugutang catalogue is bound in 32 fascicules.

If they are the same, then one would have to presume a rebinding took place in the meantime. See Jao Tsong-yi [Jao

Tsung-i] “Un inédit du Chouo-fou: Le Manuscrit de Chen Han, de la période kia-tsing (1522-1566),” in Mélanges de

Sinologie offerts à Monsieur Paul Demiéville, vol. 1 (Paris: Presses universitaires de France, 1966), p. 93; Rao

Zongyi [Jao Tsung-i] 饒宗頤, “Shuofu xin kao—Ming Jiajing Wujiang Shen Han chaoben Shuofu jilue,” 《説郛》

新考—明嘉靖吳江沈瀚鈔本《説郛》記略 in Rao Zongyi shixue lunzhu xuan 饒宗頤史學論著選 (Shanghai:

Shanghai Traditional Books Press, 1993), pp. 659. 101 See Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed., Zhongguo guji

shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989), 1.38b (p. 76 in the


籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部, pp. 1927; Shanghai guji


Shuofu sanzhong 説郛三種(Shanghai Old Binding Press, 1988), p. 1. The description in Shuofu, 1927 edition, ba 跋,

pp. 1a-b (=Shuofu sanzhong, p. 1358c-d) is, as usual, less accurate and less informative. See also the on-line

catalogue “Zhongguo guoji shanben shumu Lianhe daohang xitong” 中國古籍善本書目聯合導航系統 under 説郛

一百卷［明陶宗儀編明抄本］存五十二卷［九至十、十六至十九、二十三至三十、三十三至三十七、三

十九至四十四、五十、五十二至五十三、五十五至六十一、六十三至六十六、七十至七十二、七十九至八

十二、八十八至九十、九十二至九十四］

43

blue with one faint upper “fish tail” on the paper fold. No numbering or running header. Text written in

black ink.

History: Ming-era ms. Held at the former residence of Sun Yirang 孫詒讓 (courtesy name Zhongrong 仲

容; 1848-1908), and the fourth manuscript used by Zhang Zongxiang for the 1927 Commercial Press

edition.

Zhang 張: Currently held in the National Library of China (no. 7557). Γ recension. 91 juan extant, in 29

fascicules; the SWQZL forms juan 55 of text.102 A full table of contents has been published.103 One

preface, by Yang Weizhen 楊維楨; contents essentially identical to 1927 Commercial Press edition.104

The paper is lined in blue with one “fish tail” and white header blanks on the paper fold; however, the

pages have neither numbers nor running headers.

History: Listed as a Ming-era copy. Zhang Zongxiang attributes it to the Wanli 萬曆 period (1572-1620),

but I think it is likely to be much earlier, preceding the Fu3 copy. Jia Jingyan notes on the volume the

stamps only of the collector Zhang Yuanji 張元濟 (1867-1959)105 and the Hanfenlou 涵芬樓, i.e. the

company library of the Commercial Press, in which Zhang Yuanji was the editorial chief.106 It was the

http://202.96.31.45/libAction.do?method=goToBaseDetailByNewgid&newgid=6193&class=kind (accessed June 30,

2010). 102 Beijing tushuguan shanben bu 北京圖書館善本部, ed., Beijing tushuguan shanben shumu 北京圖書館善本書

目, vol. 5, pp. 36b-37a; Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji shanben shumu: Zi bu 北京圖

書館古籍善本書目·子部, p. 1694; Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員

會, ed., Zhongguo guji shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe,

1989), 1.38a (p. 75 in the continuous pagination added to the 1990 reprint); Weng Lianxi 翁連溪, ed. Zhongguo guji

shanben zongmu 中國古籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部,

pp. 1927. 103 Shangwu yinshuguan 商務印書館, ed., preface by Zhang Yuanji 張元濟, Hanfenlou jinyu shulu 涵芬樓燼餘書

錄 (Shanghai: Commercial Press, 1951), vol. 3, Zi 子, 57b-63a. This description contains a complete table of

contents. 104 The ms itself has only a partial table of contents covering juan 1-8 at the beginning, with no author or dynasty

attributions. 105 On him, see Manying Ip, The Life and Times of Zhang Yuanji, 1867-1959 (Beijing: Commercial Press, 1985). 106 Jia Jingyan, Shengwu qinzheng lu jiaoben 聖武親征錄校本 (unpublished ms, 乙未), I, 3b.

http://202.96.31.45/libAction.do?method=goToBaseDetailByNewgid&newgid=6193&class=kind

44

third of the mss used in the production of the 1927 Commercial Press edition, used to collate the text of

the SWQZL, whose base text was the Fu3 ms.107

Fu 傅: A set of three or four108 different fragmentary mss assembled by Fu Zengxiang 傅增湘 (1872-1950)

to make an almost complete set. Currently kept in the Shanghai Library (nos. 786660-786719). All are Γ

recensions109; the SWQZL is found in juan 55. A complete table of contents of the entire Fu ms set has

been published.110 Fu1 (juan 1-25) is on black-lined “Congshutang 叢書堂” stationary, 10 columns per

page; Fu2 (juan 26-30 and 96-100) is on blue-lined “Hongnong Yang shi 弘農楊氏” stationary, 11

columns per page; Fu3 (juan 31-67) and Fu4 (juan 68-70) are on black-lined “Shuofu 説郛” stationary, 13

columns per page.

107 This ms is the same as the “Hanfenlou 涵芬樓 ms” mentioned in Zhang Zongxiang’s colophon. Zhang

description of it in his colophon runs as follows: “One is a ms kept in the Hanfenlou library, which seems to be a

Wanli era copy and does not lack a single one of the juan. In front of each juan there is a table of contents and the

present table of contents has been copied from this ms”; see Shuofu (1927), colophon/1a; Shuofu sanzhong, p. 1358c.

This description would seem to preclude this ms being the Zhang ms, since it seems to state that it is complete and

lacks no juan. However, Zhang’s 1927 description here is misleading. In the more accurate description of the mss

used by Zhang Zongxiang given in the reprint Shuofu sanzhong 説郛三種, it is stated that this Hanfenlou ms is an

incomplete Ming ms with 91 juan, the exact number of the Zhang ms. See Shanghai guji chubanshe 上海古籍出版

社, “Shuoming 説明” in Zhang Zongxiang 張宗祥, “Shuofu jiaokan ji 説郛校勘記” in Shuofu sanzhong 説郛三種(Shanghai Old Binding Press, 1988), p. 1. 108 There is a difference of opinion between bibliographers on whether there is a separate Fu4 document. The

differences in paper and handwriting between Fu1, Fu2, and Fu3 are obvious. But Fu Zengxiang also separates out

Fu3 (juan 31-67) and Fu4 (juan 68-70) as two separate texts even though both are on very similar black lined paper

with 13 columns per page and “Shuofu 説郛” printed on the paper fold. See Mo Youzhi 莫友芝, with Fu Zengxiang

傅增湘, Cangyuan dingbu Lüting zhijian zhuanben shumu 藏園订補郘亭知見傳本書目, vol. 2, 10B/751-52. Zhang

Zongxiang, on the other, does not distinguish Fu3 and Fu4, thus seeing the Fu Zengxiang SF set as composed of

only three mss. See Shuofu (1927), colophon/1a; Shuofu sanzhong, p. 1358c and the “explanation” in Shanghai guji


Shuofu sanzhong 説郛三種(Shanghai Old Binding Press, 1988), p. 1. Between juan 67 and 68, a new hand clearly

starts, but that would not necessarily mean they derive from a different manuscript of the SF. Nor do the very sight

differences in stationary seem to me incompatible with them being simply differing printings from the same blank

stationary block-print. I thus would lean towards Zhang Zongxiang’s viewpoint on this issue. 109 Zhang Zongxiang notes in his colophon to his printed edition that “the numbering of the juan has some

discrepancies with the table of contents,” but this must refer to only the very minor discrepancies that can likewise

be found between the table of contents and the actual text in the Zhao, Zhang, and other mss. 110 See Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed., Zhongguo guji

shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989), 1.25a-38a (pp. 49-

75 in the continuous pagination added to the 1990 reprint), and Weng Lianxi 翁連溪, ed. Zhongguo guji shanben

zongmu 中國古籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部, pp.

1923-1927.

45

History: Fu1 is written on stationary of the Congshutang 叢書堂 library of Wu Kuan 吳寬 (1435-1504).

Fu2 is dated to Fu3 has a note in juan 62: “completed in the third month of Hongzhi 18” or A.D. 1505.

The three (or perhaps four) mss were brought together by Fu Zengxiang as marked by his Shuangjianlou

雙鑒樓 stamp at the beginning of Fu1. Stamps of Wang Tiren 王體仁 (courtesy name Shoushan 綬珊,

1873-1938)111 and the Shanghai Library stamp found periodically throughout the set. This was the second

of the mss used by Zhang Zongxiang in the 1927 Commercial Press edition. The text of Fu3 was used as

the base text for that edition’s SWQZL.

Uang 汪: Currently held in the Zhejiang Library (no. 7434). Γ recension. 41 juan extant in 26 fascicules;

SWQZL found in juan 55.112 A full table of contents has been published.113 Juan with preface(s) and/or

contents missing. The organization has a number of differences in detail with the 1927 Commercial Press

edition. The paper is lined in light blue, with 9 columns per page and 24 characters per column. The page

fold has one “fishtail” and a header blank. Running header with juan no. and page numbers throughout.

History: Ming era copy. Stamps of Wang Wenbo 汪文柏 (courtesy name Jiqing 季青, sobriquet Keting

柯庭), active in the Kangxi era (1662-1722), originally of Xiuning 休寧 (Anhui), later of Tongxiang 桐鄉

(Zhejiang).114 Zhang Zongxiang located this copy in the Zhejiang library in 1952 and used it to collate his

published 1927 edition of the Shuofu, which notes the editors of Shuofu sanzhong reproduced in an

appendix.115

111 The seal reads: Hangzhou Wang shi Jiufengjiulu cangshu zhi zhang 杭州王氏九峰舊廬藏書之章. 112 The catalogue entry in must be the same manuscript said to contain 45 juan 卷 found in Zhongguo guji shanben

shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed., Zhongguo guji shanben shumu: Congbu 中國古籍

善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989), p. 1.38b (p.76 in the continuous pagination added to

the 1990 reprint), and in Weng Lianxi 翁連溪, ed. Zhongguo guji shanben zongmu 中國古籍善本總目 (Beijing:

Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部, p. 1927. The list of extant juan, however, does

not at all match what is in the Zhejiang library catalogue (see below). 113 Zhejiang Tushuguan guji bu 浙江圖書館古籍部, ed., Zhejiang Tushuanguan guji shanben shumu 浙江圖書館古

籍善本數目 (Hangzhou: Zhejiang Educational Press, 2002), pp. 651-54. 114 His stamps read Xiūníng Wāng Jìqīng jiā cáng shūjí 休寧汪季青家藏書籍 and Gǔ xiāng lóu 古香樓. 115 See Shanghai guji chubanshe 上海古籍出版社, “Shuoming 説明” in Zhang Zongxiang 張宗祥, “Shuofu jiaokan

ji 説郛校勘記” in Shuofu sanzhong 説郛三種(Shanghai Old Binding Press, 1988), pp. 1ff.

46

Hū 滹: Currently held by National Library of China (no. 0485).116 Δ recension. 55 juan extant in 17

fascicules; does not include the SWQZL. Table of contents and preface(s) not preserved, but a later owner

wrote in the contents of each fascicule on the reverse of its cover page. Selected contents listed in Table 2.

Paper lined in orchid (lan 蘭) with 13 columns per page and 19-20 characters per column. The page fold

has two “fishtails” defining two header blanks. The lower one has the running header Hūnan shushe 滹南

書舍. Single line border on all four sides.

History: may be related to the Yu Wenbo 郁文博 recension completed in 1496. Copied by the Hūnan

Printing House (Hunan shushe 滹南書舍) on to their own stationary. They or a subsequent owner

addednumerous proof-readers marks (〇 and 丶), as well as annotations in the top margins and

corrections between the lines.117 Commented on by Fu Zengxiang.118

Shi 史: Currently held in the National Library of China, as “Shuofu not divided into juàn” (no.

A01507).119 Ε recension. 12 fascicules with contents equivalent to roughly 20 juan. No juan

numbering.120 Contents listed in Table 4. Paper lined in blue with 14 columns per page and an absolutely

consistent 22 characters per column. The mise en page is identical to that of the other ε recension ms,

116 See Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji shanben shumu: Zi bu 北京圖書館古籍善本書

目·子部, p. 1694. Cf. Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed.,

Zhongguo guji shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989),

1.38a (p. 75 in the continuous pagination added to the 1990 reprint); Weng Lianxi 翁連溪, ed. Zhongguo guji

shanben zongmu 中國古籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部,

pp. 1927. I made a cursory examination of a microfilm of this ms in June, 2009, and longers ones in July, 2012, and

December, 2013. 117 Since I have only had access to the black and white microfilm, I cannot tell if these annotations are, as one would

expect, in red ink. 118 Mo Youzhi, Cangyuan dingbu Lüting zhijian zhuanben shumu, p. 752 (second Shuofu listed). 119 Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji shanben shumu: Zi bu 北京圖書館古籍善本書目·子

部, p. 1695; Zhongguo guji shanben shumu bianji weiyuanhui 中國古籍善本書目編輯委員會, ed., Zhongguo guji

shanben shumu: Congbu 中國古籍善本書目·叢部 (Shanghai: Shanghai guji chubanshe, 1989), 1.38b (p. 76 in the


籍善本總目 (Beijing: Xianzhuang shuju, 2005), vol. 6, Ji bu (xia), cong bu 集部(下), 叢部, pp. 1927. 120 A later curator of the ms occasionally added in numbers. Thus with Daye zaji 大業雜記, the number 57 is written

in. But this is derived not from counting up the juan in the actual Shi ms, but from the numbering in the γ recension.

47

Shen. It has white header blanks and a single border on all four sides. It has a single fish-tail and a

running header of SF 説郛, but no page numbers. 121

History: Ming-era copy, with text superior to the Jiajing-era Shen ms. Stamp of Jieshushanfang Studio 借

樹山房 of the scholar Shi Mengjiao 史夢蛟 (Qianlong era).122 A number of marginal notes or comments

of three types: 1) proofreaders corrections, found in the earlier five fascicules; 2) juan nos., derived from

the γ recension, pasted in or written in pen; 3) notes by a scholar with the courtesy name Xiaozheng 曉鉦,

dated to year ding/si 丁巳 whom Jia identified with Qian Daxin 錢大昕 (courtesy name Xiaozheng 曉徵)

and ding/si (Jiaqing 2, A.D. 1798). Acquired by National Library of China in the Republican period.123

Shen 沈: Currently held in the Fung Ping Shan Library of the University of Hong Kong (cat. no. 善

837/77-11).124 Ε recension, complete in 69 juan, in 24 fascicules; the SWQZL forms 60th juan. Described

by Jao Tsung-i, with contents of representative juan. 125 No prefaces, table of contents, or juan

numbers.126 Four colophons written by Lu Qiao 陸樵 (see below). Written on white tissue paper, lined in

121 Based on my personal examination of the microfilm and Beijing tushuguan 北京圖書館, ed., Beijing tushuguan

guji shanben shumu: Zi bu 北京圖書館古籍善本書目·子部, p. 1695. 122 See also Jia, I, zhuiyan, p. 4a-b; Beijing tushuguan 北京圖書館, ed., Beijing tushuguan guji shanben shumu: Zi

bu 北京圖書館古籍善本書目·子部, p. 1695. I know Shi Mengjiao only as the publisher of the complete works of

the Ningbo scholar Quan Zuwang 全祖望 (1705-1755) entitled Jieqiting quanji 鮚奇亭全集 and a chronology of

Quan’s life Qing Quan Xieshan xiansheng Zuwang nianpu 清全謝山先生祖望年譜. 123 The library stamp reads Guoli Beiping tushuguan suo cang 國立北平圖書館所藏. 124 See the Fung Ping Shan Library online catalogue at http://bamboo.lib.hku.hk/fpslindex/full_list.asp?RID+721 . 125 See Jao Tsong-yi [Jao Tsung-i] “Un inédit du Chouo-fou: Le Manuscrit de Chen Han, de la période kia-tsing

(1522-1566),” in Mélanges de Sinologie offerts à Monsieur Paul Demiéville, vol. 1 (Paris: Presses universitaires de

France, 1966), pp. 87-104; Rao Zongyi [Jao Tsung-i] 饒宗頤, “Shuofu xin kao—Ming Jiajing Wujiang Shen Han

chaoben Shuofu jilue,” 《説郛》新考—明嘉靖吳江沈瀚鈔本《説郛》記略 in Rao Zongyi shixue lunzhu xuan 饒

宗頤史學論著選 (Shanghai: Shanghai Traditional Books Press, 1993), pp. 654-666; and Rao Zongyi [Jao Tsung-i]

饒宗頤, Xianggang daxue Feng Pingshan tushuguan cang shanben shulu 香港大學馮平山圖書館藏善本書錄

(Hong Kong: Lungmen Bookstore, 1970), pp. 158-164. 126 As Jao notes in a footnote to his French article, “Un inédit du Chouo-fou,” p. 90n.1, there are in fact four places

where the juan no. is noted, but the numbering is not consistent with the current organization. Thus fascicule 20 has

one juan labeled no. 6, fascicule 21 has one juan labeled no. 15, fascicule 22 has one juan labeled no. 40, and

fascicule 24 has one juan labeled no. 40. Since fascicule 24 is the last one, and the Šn ms has a total of 69 ms, it

seems clear that some fascicules which were originally near the end of the work have been moved towards the

beginning of it. A similar phenomenon appears in the Shi ms.

http://bamboo.lib.hku.hk/fpslindex/full_list.asp?RID+721

48

black with 14 columns per page and an absolutely consistent 22 characters per column. The mise en page

is identical to that of the other ε recension ms, Shi. The page fold has three “fish tails” and a running

header of Shen 沈.

History: Jiajing-era ms. As noted by Jao Tsung-i, the stationary and stamp belong to Shen Han (courtesy

name Yuanyue 原約 , from Wujiang 吳江) and post-date his jinshi 進士 degree in 1535. Subsequent

owners include Huang Jishui 黄姬水 (1509-1574) and Lu Qiao 陸樵, who added colophons dating his

acquisition of the ms to ji/chou 己丑, probably 1589.127 Later owners include Lù Yunxiang 陸雲祥

(courtesy name Jiaqing 嘉卿, juren 舉人 degree 1627) of Wujiang 吳江, Lú Zhi 盧址 (1725-1794) in his

Baojinglou 抱經樓 Residence, and Liu Chenggan 劉承幹 (courtesy name Zhenyi 貞一, sobriquet Hanyi

翰怡, 1881-1963) of Nanxun 南潯, before being acquired by the Fung Ping Shan library.

Han draft 涵稿: Currently held in the Zhejiang Library (no. 7437). Γ recension. All 100 juan extant in 52

fascicules; the SWQZL forms juan 55 of the text. A complete table of contents has been published.128 This

ms is the draft for Zhang Zongxiang’s 1927 Commercial Press edition, designated by Jia as the Hanfenlou

or Han 涵 edition. Prefaces and contents are thus identical with those of the printed edition. The base text

for each work is written in black ink with corrections, usually based on some other manuscript, but in

some cases based only on Zhang’s editorial judgment, written in in red ink. The black ink text thus

represents a copy of the manuscripts used by Zhang, i.e. the extant Fu, Zhang, and Sun mss and the now

lost Metropolitan ms. Thus although for most juan it is of interest solely for understanding Zhang

Zongxiang’s editorial process, for those juan, where he his base text was the lost Metropolitan text, this

forms its only extant copy.

127 Although Jao Tsung-yi did not present any additional information on Lù Qiao, given that the ms was certainly

produced after 1535, and was in Lú Zhi’s possession by the Qianlong era, only dates of 1589, 1649, and 1709 are

possible. The colophon’s criticism of the 100-juan ms as the most current one and the absence of reference to the

120-juan late-Ming-early Qing printed edition would seem to exclude 1649 or 1709, leaving 1589 as the only

possibility. 128 Zhejiang Tushuguan guji bu 浙江圖書館古籍部, ed., Zhejiang Tushuanguan guji shanben shumu 浙江圖書館古

籍善本數目 (Hangzhou: Zhejiang Educational Press, 2002), pp. 670-80.

49

Taipei 台: Currently held in the Central National Library (Guoli zhongyang tushuguan 國立中央圖書館)

in Taipei (no. 000525628). Γ recension. All 100 juan extant in 64 fascicules; the SWQZL forms juan 55 of

text. A complete table of contents has been published.129 One preface, by Yang Weizhen 楊維楨, and

prefatory statement that text is based on Tao Zongyi’s as reorganized “Du” (error for Yu) Wenbo 都

(error for 郁)文博.130 Contents are very similar to those of the 100-juan SF published by Zhang

Zongxiang 張宗祥 in 1927, and Zhang ms. Each fascicule begins with separate table of contents. Paper is

lined in blue with 11 columns per page with no “fish tails” and no running header. The size is 18.6 x 14.1

centimeters.

History: Forgery, post-dating 1926. The SWQZL text created by copying the Fu3 manuscript text and then

collating it with Wang Guowei’s 1926 scholarly edition. This collation and the addition of the statement

of Yu Wenbo 郁(~都)文博 and the omission of the clearly Yongle-era text in juan 97 probably related to

the attempt to give an appearance of a highly valuable text.

Currently Unlocated Shuofu mss (all γ recension mss) Metropolitan 京: A manuscript used by Zhang Zongxiang for his 1927 edition of the Shuofu. (This should

not be confused with the Fu-Metropolitan ms, called the Metropolitan ms by Jia Jingyan.) His description

of it states that it was kept in the Metropolitan Library (京師圖書館):

One [exemplar used] is a fragmentary edition in the Metropolitan Library (juàn no. 3, no. 4, and

nos. 23-32). This has no year dating, and is written on white tissue paper; the calligraphy is

extremely big and tall. It seems to be a ms copied during the Longqing 隆慶-Wanli era. 131

The holdings of the Metropolitan Library were inherited by the National Library of China, so

presumably it should be extant. However it is not listed in any catalogue known to me. However,

129 Guoli zhongyang tushuguan shanben shumu 國立中央圖書館善本書目, 2nd edition, vol. 4, pp. 1445-84, esp. p.

1470. 130 See the page reproduced in Chang Bide, Shuofu kao, pl. 1. 131 Shuofu (1927), colophon/1a; Shuofu sanzhong, p. 1358c.

50

if Zhang Zongxiang did indeed use this ms as his base text for juan 3, 4, and 23-32, then the

black ink text of the Han draft manuscript for those juan ought to be a more or less accurate

copy of the Metropolitan ms. Further stemmatic analysis could then identify the position of this

ms in the stemma.

She 涉: A manuscript described by Wang Guowei as being of the Wanli era, and owned by Tao Xiang 陶

湘 (courtesy name Lanquan 蘭泉, sobriquet Sheyuan 涉園, 1870-1940), from Wujin 武進. Wang Guowei

visited Tao Xiang in Tianjin and borrowed the ms, using it to collate his edition of the SWQZL.132 Jia

Jingyan also refers to the Sheyuan 涉園 or She 涉 ms, but instead of using it directly, he used a copy of

the 1901 Japanese reprint of the He Qiutao edition of the SWQZL which had Wang’s notes in it, kept in

the National Library of China.133 Neither Wang nor Jia made much use of this edition, seeing it as

essentially identical to Fu3. Indeed my analysis of their collations shows that its text of the SWQZL is

likely a codex descriptus, identical to Fu3. Given the description of Wang, who emphasizes its similarity

to the Fu ms, it may be assumed to be a γ recension text.

Fu-Metropolitan ms 傅京: This ms is described by Jia Jingyan as a Ming-era copy of the SF, which is

copied onto stationary with the running header “Copy from the Metropolitan Library” (Jingshi tushuguan

chao 京師圖書館鈔), hence its name. Since the Metropolitan Library existed only from 1909 to 1928,

what Jia must mean is that it is a recent copy of a Ming-era ms. It is, he says, currently kept in the

National Library of China, although as a twentieth century ms it was evidently not included in the

catalogues of rare books and mss of the National Library.134 My stemmatic analysis shows that its text of

the SWQZL is a copy of the Fu3 ms that incorporates a small number of editorial emendations, some

132 Wang Guowei 王國維, “Shengwu qinzheng lu 聖武親征錄,” in Menggu shiliao sizhong 蒙古史料四種 (Taiwan:

Cheng-chung Press, 1962), p. 1b (2). 133 Jia, I, Zhuiyuan, 5a. This volume is not listed in the catalogues of rare books in the National Library of China,

presumably because the text in which Wang made his notes was not the 1894 Chinese edition, but the 1901 Japanese

reprint. 134 Jia, I, pp. 4a, 5b.

51

derived from the then most current edition of the SWQZL, that of He Qiutao. Location of this ms would

assist in understanding Zhang Zongxiang’s research on the Shuofu but would not have any significance

for Shuofu ms studies.

Yue 粵 ms: In May of jia/wu 甲午 (1951), Zhang Zongxiang made a collation of a ms from Guangzhou’s

Yueyatang 粵雅堂 traditional publishing house against his 1927 printed edition, before making another

collation against the Uang ms. These two collations were published together by the editors of the Shuofu

sanzhong, although the editors were unable to distinguish the notes pertaining to the Yueyatang 粵雅堂

ms from those pertaining to the Uang ms. Thus what they reprinted is simply a list of all the alternative

readings from these two collations, along with his own editorial notes. The Yueyatang ms has not, to my

knowledge, been identified yet. Since, however, the Uang ms is extant, presumably comparison of all the

collations given in Shuofu sanzhong with the Uang ms would enable one to exclude Uang readings, thus

leaving only the Yueyatang readings, which could then be used to search for this ms. (It is also possible

that the Yue ms is in fact written on Yueyatang stationary, which would make its identification much

simpler.) Indeed, the collations listed by the editors included those in juan 8, 19, and 93, which are not

extant in the Uang Shuofu, at least according to the published table of contents. Thus those collations are

likely to be from the Yue ms.

52

Tables

Table 1: The addition and loss of works in the Shuofu, illustrated by a sample of works on Northern,

overseas, and border topics.

Α recension = Mao ms. of 1361; γ recension = Zo, N, Zg, and T mss and 1927 Commercial Press Edition;

ζ recension = late Ming print (as kept in Kyoto Institute of Oriental Culture and reprinted in Shuofu

sanzhong).

Topic

Author and era

Name

α

γ

ζ

Kitan Liao

— (Song) 宋闕名 Shi Liao lu 使遼録 48 3 —

Wu Gui (Song) 宋武珪 Yanbei zaji 燕北雜記 40 4 50

Yang Boyan (Song) 宋楊伯

喦

Yi cheng (~jian) 臆柬(~乘) 9 21 11

Wang Yi (Song) 宋王易 Chongbian Yanbei lu 重編燕北録 — 38 56

Ye Longli (Song) 宋葉隆禮 Liao zhi (abridged) 遼志 — 86 55

Hu Jiao (Five Dynasties) 五

代胡嶠

Xianlu ji 陷虜記 — — 56

Jurchen Jin

Hong Hao (~Mai) (Song) 宋

洪皓(~邁)

Songmo jiwen 松漠紀聞 29 8 55

Wen Weijian (Song) 宋文惟

簡

Luting shishi 虜廷事實 3 8 55

— (Song) 宋闕名 Beifeng yangsha lu 北風揚沙録 49 25 55

Cheng Dachang (Song) 宋程

大昌

Beibian beidui 北邊備對 — 52 56

Zhou Hui (Song) 宋周煇 Bei yuan lu 北轅録 — 54 56

Yuwen Maozhao (Song) 宋

宇文懋昭

Jinguo zhi (abridged) 金國志 — 86 55

Shi Maoliang (Song) 宋石茂

良

Birong ye (~jia) hua 避戎夜(~嘉)

話

— — 37

Mongol

Yuan

Meng/Zhao Gong (Song) 宋

孟(for 趙)珙

Meng-Da beilu 蒙韃備録 — 54 56

— (Yuan) 元闕名 Shengwu qinzheng lu 聖武親征録 — 55 —

Liu Yu (Yuan) 元劉郁 Xishiji 西使記 — — 56

Korea

Sun Mu (Song) 宋孫穆 Jilin leishi 雞林類事 31 7 55

Fang Feng (Song) 宋方鳳 Yisu kao 夷俗考 — — 55

Xu Jing (Song) 宋松兢 Shi Gaoli lu 使髙麗録 — — 56

Vietnam Xu Mingshan (Yuan) 元徐明

善

Annan xingji 安南行記

—

51

56

Cambodia Zhou Daguan (Yuan) 元周達

觀

Zhenla fengtuji 真臘風土記

—

39

62

53

Burma — (Tang) 唐闕名 Piaoguo yuesong 驃國樂頌 — 67 100

Yunnan

Li Jing (Yuan) 元李京

Yunnan zhilue 雲南志略

—

36

62

Manchuria

Qi Fuzhi (Yuan) 元戚輔之

Liaodong zhilue 遼東志略

—

97

62

Kökenuur

Li Yuan (Song) 宋李遠

Qingtang lu 青塘録

—

35

—

Uyghuristan

Wang Yande (Song) 王延德

Gaochang jixing 髙昌行紀

—

—

56

Sources: Xu Sanjian, “Jiguge cang Mingchao liushijuan ben ‘Shuofu’ kao shu,” Dongnan wenhua 1994,

no. 6 (no. 106), pp. 118-27; personal examination of mss in National Library of China, nos. 2408, 3907;

Shangwu yinshuguan, ed., Hanfenlou jinyu shulu (Shanghai: Commercial Press, 1951), vol. 3, Zi 子, 57b-

63a; Guoli zhongyang tushuguan shanben shumu, 2nd edition, vol. 4, pp. 1445-84; Chang Bide, Shuofu

kao (2nd edition, Taipei: Wen-shih-che Publishing House, 1979), 43-405, 483-506; Shuofu sanzhong

(Shanghai: Shanghai Old Binding Press, 1988), 10 vols.; Tōhō bunka gakuin Kyōtō kenkyūsho kanseki

mokuroku (Kyoto: Tōhō bunka gakuin Kyōtō kenkyūsho, 1938), pp. 324-47.

54

Table 2: Contents of the Hu ms Shuofu compared to the standard γ recension Hu ms juàn Equivalent juàn in the γ recension Specific Works (selected)

6 3 談壘 (pt. 2), 古杭夢游錄

7 4 (first half) 墨娥漫錄 (pt. 1 to 仇池筆記)

8 4 (second half) 墨娥漫錄 (pt. 2 to 封氏聞見記)

9 5 (first half) 鶴林玉露

10 5 (second half) 傳載

11 6 讀子隨識

15 8 (first half) 玉澗雜書

16 8 (second half) 捫蝨新話

21 11 玉泉子眞錄

22 12 (first half) 悅生隨抄

23 14 (end part),

12 (second half) 博異志

洞天淸錄集

24 13 書鑒

25 14 就日錄

26 15 因話錄

27 16 (first half) 三器圖義

28 16 (second half) 雲林石譜, 宣和石譜

29 17 (second half) 愛日齋叢鈔

30 18 (first half) 坦齋筆衡

31 18 (second half) 碧雞漫志

32 19 (first half) 打馬圖經

33 19 (second half) 甘澤謠

34 20 (middle part) 儒林公議

35 20 (end part) 植跋簡談

36 21 (end part) 昨夢錄

37 22 (second half) 山家清供

38 23 (first half) 賔退録

39 23-24 諧史, 麈史, 歸田録, 孔氏雜説, 湘

山野録, 逸更 [sic]

40 24 (second half) 墨客揮犀, 肯綮録

41

75, 25 土林紀實, 卓異記, 集異記, 桐譜

42 26 宣政雜録, 洛陽名園記

64 39 (middle) 陶朱新録, 真蠟風土記

68 40 慎子

69 44 靖康朝野僉言

70 84 錢譜

71 74 (beginning and end parts) 褚氏遺書, 大事記, 白虎通德論, 辨

惑論

72 74 (middle) 大中遺事

73 47 公孫龍子

74 48 聱隅子歔欷瑣微論

75 50 識遺

76 51 豫章古今記, 安南行記

77 52 北邊備對

78 53 (first half) 鉤玄

55

79 53 (second half) 四朝聞見雜録

80 97 金山志, 遼東志, 稽古定制, 勸善録,

夷堅志, 神僧傳, 效顰集

88 58 江表志

89 60 品茶要録

90 59 (first half) 史記注語, part 1

91 59 (second half) 史記注語, part 2

56

Table 3: The Shen ms of the Shuofu in comparison with the Mao and γ recension mss

Work in Shen ms

Mao ms

毛本

γ recension mss

Xue dao xuan zhen jing 學道玄真經 Vol. no. 1

第一册

— 54

Gan ying Jing 感應經 16 9

Yang yu jing 養魚經 15 15

Xiang he jing 相鶴經 15 15

Xiang ju (~bei) jing 相具（貝）經 15 15

Tu niu jing 土牛經 15 15

Da ma tu jing 打馬圖經 16 19

Jiu jing 酒經 — 44

Du bei shan jiu jing 讀北山酒經 — 44

Zui xiang ri yue 醉鄉日月 — 58

Pin cha yao lu 品茶要録 One juàn

— 60

Xuanhe bei yuan gong cha lu 宣和北

苑貢茶録

— 60

Bei yuan bie lu 北苑別録 — 60

Da guan cha lun 大觀茶論 — 52

Mo e man lu 墨娥漫録 One juàn 40 4

Feng tu ji 風土記. . . 40 4

Chouchi biji 仇池筆記 — 4

Zhu zi sui shi (128 entries) 諸子隨識 Vol. no. 8

第八册

卷?

35 6

Wen zi 文子. . . 35 6

Yin wen zi 尹文子. . . 35 6

Huai nan zi 淮南子 . . . .

35 6

Lun heng 論衡

— 100

Zhu zhuan zhai xuan 諸傳摘玄

Vol. no. 9

第九册

卷?

36 7

Gao seng zhuan 髙僧傳

36 7

Wuming gong zhuan 無名公傳

— 73

57

Xie lue 蟹略 Final juàn

末卷（第幾號?）

— 36

Zhu yi feng su 諸夷風俗 — —

Zhenla feng tu ji 真臘風土記 — 39

Sources: Rao Zongyi, “‘Shuofu’ xin kao—Ming Jiajing Wujiang Shen Han chaoben ‘Shuofu’ jilue,” in

Rao Zongyi shixue lunzhu xuan (Shanghai: Shanghai Traditional Books Press, 1993), p. 661; Xu Sanjian,

“Jiguge cang Mingchao liushijuan ben ‘Shuofu’ kao shu,” Dongnan wenhua 1994, no. 6 (no. 106), pp.

118-27; Chang Bide, Shuofu kao (2nd edition, Taipei: Wen-shih-che Publishing House, 1979), 43-405,

483-506.

58

Table 4: Contents of the Shi ms of the Shuofu in comparison with the γ recension mss.

Fascicule Title Juan in γ recension

I. 傳載 5

藏一話腴 5

墨客揮犀 24

續墨客揮犀 24

藝圃折中 31

II 讀子隨識 6

鬼谷子三卷 71

亢倉子 71

鬼谷子五卷 71

III 迷樓記 32

教坊記 12

卓異記 25

集異記 25

IV 。。。柭印斗秤 ?

趨朝事類 34

麟臺故事 34

北邊備對 52

V 省心詮要 35

甘澤謠 19

鐵圍山叢談 19

韋居聽輿 21

白獺髓 25

三水小牘 33

羣居解頤 31

鉤玄 53

稽古定制 97

VI 遼志 86

遼東志略 97

金國志 86

雲南志略 36

VII 煬帝開河記 44

…..

VIII 墨子 46

子華子 46

曾子 46

尹文子 46

孔叢子 46

IX 。。。萬機論 ?

素書 90

聱隅子歔欷瑣微論 48

韓非子 47

X 北轅錄 54

蒙韃備錄 54

虜庭事實 8

溪蠻叢笑 5

59

XI 。。。長城記 ?

聖武親征錄 55

XII 大業雜記 57

嶺表錄異記 34

海山記 32

《説郛》版本史 ("Textual History of Tao Zongyi's Shuofu")

Documents

Transcript of 《説郛》版本史 ("Textual History of Tao Zongyi's Shuofu")