Editorial preface · the 1847 memorandum and this bilingual edition
编者前言 · 1847 年备忘录及本双语版本
What Mommsen sent to Berlin, and what this edition adds
蒙森呈交柏林之原稿，与本版所增之注
In January 1847, a thirty-year-old jurist sat in a Roman lodging and wrote out a thirty-two-page memorandum addressed not to the public but to the seventy-odd members of the Royal Prussian Academy of Sciences in Berlin. Theodor Mommsen (1817–1903) had just spent two years travelling Italy on the Academy's stipend, transcribing inscriptions in libraries, in museum storerooms, and at findspots from the Po valley down to Calabria. He held no permanent post; the cover identifies him only as Doctor der Rechte, a doctor of laws. The Academy that had funded him was, however, the same body that in 1815 had commissioned August Böckh's Corpus Inscriptionum Graecarum, a project still half a century from completion. A string of parallel Latin attempts had recently failed: Olaus Kellermann's in Rome, the Paris project under Letronne. The cover read "Gedruckt als Handschrift für die Herren Mitglieder der Königl. Akademie der Wissenschaften zu Berlin", that is, printed as manuscript for the Academy's members. It was not a publication; it was an internal position paper, and its argument was that the Berlin Academy could succeed where everyone else had failed.
1847 年 1 月，一位三十岁的法学博士在罗马一处寓所里写就一份三十二页的备忘录。他叫蒙森 (Theodor Mommsen, 1817–1903)，刚结束由柏林皇家普鲁士科学院 (Königliche Preußische Akademie der Wissenschaften) 资助的两年意大利之行，从波河谷 (Po valley) 一直考察到南端的卡拉布里亚 (Calabria)，把图书馆、博物馆库房、田野里能见到的拉丁铭文都抄了一遍。他没有固定教职，封面上只自署 Doctor der Rechte，法学博士而已。然而资助他的那所学院，正是 1815 年委托勃克 (August Böckh) 编纂《希腊铭文大全》(Corpus Inscriptionum Graecarum) 的同一个机构；而那部书直到 1877 年才告竣。与此同时，欧洲一连串建立平行拉丁语料的尝试相继失败：罗马的 Kellermann 项目、Letronne 在巴黎主持的项目，都未能成事。1847 年这一年，约莫是我国清道光二十七年，魏源《海国图志》初版方出不久，龚自珍则刚逝世六年。中国学界正在重新打量"舆地之学"如何在新的方法下展开，而柏林这边亦正在思考同样的问题，只不过对象换成了罗马的石头。封面那行 "Gedruckt als Handschrift für die Herren Mitglieder der Königl. Akademie der Wissenschaften zu Berlin"（"以手稿形式印行，专供皇家普鲁士科学院诸位院士"）已经把这份文件的性质说清楚了：它不是公开出版物，而是一份内部立场陈述，其论点正是：柏林学院可以在他人皆败之处成事。
Read in retrospect, the memorandum is the founding charter of modern epigraphy, and arguably of the modern, multi-decade, state-funded humanistic Großbetrieb ("large-scale enterprise") in general. Mommsen condenses the entire programme into a single sentence on page 3: collect all Latin inscriptions; order them conveniently; after excising the forgeries, edit them critically from the latest accessible sources, marking variant readings; provide precise indices. Commentary is desirable, but not necessary. Each of the four clauses was a leap. "Collect all" set an exhaustive corpus against the convention of selective florilegia. "Order conveniently" turned out to mean topographically (by findspot) rather than systematically (by content), and the argument was structural rather than aesthetic: only a geographically arranged corpus absorbs new finds without re-shuffling its earlier entries. "Edit critically from the latest sources" made autopsia, the personal inspection of the stones, not a virtue but a threshold; the word Mommsen actually chose was erforderlich ("required"), not wünschenswerth ("desirable"). "Precise indices" committed the corpus to discoverability, that is, to what we would now call a queryable database. The pragmatic move sat in the last clause: by relegating commentary to "desirable but not necessary," Mommsen kept the project finite.
若回头看，这份备忘录其实就是现代铭文学的奠基章程；甚至可以说，它也是现代意义上那种跨数十年、由国家资助的人文"大企业" (Großbetrieb) 的奠基章程。蒙森在第 3 页把整套纲领凝缩成一句：将全部拉丁铭文汇为一集；以便利之次序排列；剔除伪刻之后，依尽可能取自最末端可达之来源所成之文本，附重要异读，作批判性精确再现；并以精确之索引便利使用。评注虽属可取，然非必需。这四条，每一条都不是小动作。"全部收录"，是要把穷尽性总集摆到选辑性文存 (florilegia) 这一传统的对面去。"以便利次序排列"，看起来含糊，实际上指的是地形式 (按出土地) 而非系统式 (按内容)；蒙森的论据并非美学，而是结构性的。真正的问题恐怕是：一部按内容分类的总集是闭合的，按地理排列的总集才是开放的，唯有后者能在不重排既有条目的前提下吸纳新发现。"依最末端来源作批判性版本"，把亲勘 (autopsia)，即亲眼勘察石头，从博尔盖西 (Bartolomeo Borghesi) 个人的好习惯升格为制度门槛；蒙森选用的词是 erforderlich（必需），不是 wünschenswerth（可取），一字之差，差别极大。"精确索引"使总集致力于可发现性，亦即今日所谓"可查询的数据库"。真正务实的一笔，却落在最后一句：把评注降为"可取而非必需"，看似谦退，实则是使整个工程在有限的时间内得以完成的关键决断。蒙森并非没有意识到评注有学术价值，他只是看清了一点：与其面面俱到而项目永远做不完，不如先把基础建好，把评注留给后世。
The rest of the story is the subject of the "After 1847" section that closes this page. The Berlin Academy approved the plan in 1853, the first CIL volume appeared in 1863, the Italian volumes containing CIL X 7296 in 1883, and by Mommsen's death in 1903 the project was substantially complete. Mommsen's organisational model, what Chad Wellmon (following Adolf Harnack) calls the Großbetrieb der Wissenschaften or "large-scale production of the sciences," became the template for every multi-institutional scholarly corpus that followed: the IG (1903–), the CIS (1881), the digital epigraphic databases of the 1990s (EDH, EDR, EDCS), the open-access EpiDoc-Aphrodisias project of 2001, the I.Sicily corpus, and most recently the Ithaca (2022) and Aeneas (2025) neural-network restoration tools trained on those very corpora. Each of these inherits one or more clauses of the 1847 programme; some have begun to fail certain clauses. Tracing those inheritances and failures is what the nine subsections of "After 1847" do.
其后发生了什么？这是本页结尾"1847 之后"一节的主题。简言之：柏林学院 1853 年批准计划，1863 年首卷面世，1883 年载有 CIL X 7296 的意大利诸卷出版，至 1903 年蒙森辞世时整个项目已大体竣事。蒙森的组织模式，由哈尔纳克 (Adolf Harnack) 首先命名、Chad Wellmon 在 2019 年加以概念化的"学术之大企业经营" (Großbetrieb der Wissenschaften)，后来成为其后所有跨机构学术总集的范本：1903 年起的 IG；1881 年起的 CIS；1990 年代的数字铭文数据库（EDH、EDR、EDCS）；2001 年开始的开放访问 EpiDoc-Aphrodisias 项目；I.Sicily 语料；以及最近以这些语料训练的神经网络恢复工具 Ithaca (2022) 与 Aeneas (2025)。这些后继者，每一个都继承了 1847 纲领的某一条，但也有几条到了它们手上开始失灵。真正值得追问的，与其说是 CIL 是否"成功"了，不如说是：这套 19 世纪德意志学院里设想出来的工作方式，在数字时代究竟在哪几个点上仍然成立，又在哪几个点上已经走到了它自己设下的限度。追溯这些继承与失灵，正是本页后半部所做的事。
How this edition is laid out. The parchment block below carries the 1847 German verbatim, with the original page numbers in the margin and a paragraph-by-paragraph Chinese translation alongside (marked 译). In a few places Mommsen argues from controversies his Berlin readers would have known but we no longer do — Borghesi's autopsy maxim, the Sarti monopoly on Vatican access, the Kellermann collapse, the topographic-versus-systematic debate. For each of these, this edition adds an editor's note in a separate cream panel linking his choice to later developments: the Leiden Conventions of 1931, probato dolo and the partial rehabilitation of Ligorio in 21st-century scholarship, EDH's classification scheme, EpiDoc's machine-readable encoding of editorial certainty. The notes stay outside Mommsen's prose, in a different typeface and colour. Read the parchment block for Mommsen's 1847 text alone; read parchment plus notes for what his decisions led to over the next 180 years.
本版的版式说明。下方羊皮纸 (parchment) 色块为 1847 年德文原稿, 逐字保留, 按原印本标页码, 每段之后配一段汉译 (标"译"字)。蒙森论证里有几处依赖的争论 (博尔盖西的亲勘格言、Sarti 对梵蒂冈馆藏的垄断、Kellermann 项目的崩溃、地形式与系统式的取舍), 当时的柏林听众熟悉, 今天的读者已陌生; 这些地方加入编者札记, 置于另一独立的米色面板, 把蒙森的判断与他当时未能预见的后续相连: 1931 年的莱顿规约、21 世纪学界在 probato dolo 原则下对利戈里奥 (Ligorio) 的部分平反、EDH 的分类方案、EpiDoc 把编辑确信度编成机器可读的属性。札记不进入蒙森的散文本身, 字体与色块都另作区分。只读羊皮纸, 看到的是 1847 年蒙森写下的话; 羊皮纸加札记同读, 看到的是他的决定在之后 180 年间的延展。
Mommsen 1847 · German original
Source: 1847 mommsen über plan und ausführung … .txt
[1]
Die Nothwendigkeit einer Sammlung aller lateinischen Inschriften braucht nicht erst bewiesen zu werden. Notorisch ist die gruter-scaligersche veraltet, die muratorische von Haus aus mangelhaft und unzulänglich, die Masse der außerhalb der beiden Hauptsammlungen befindlichen, in hundert Büchern zerstreuten Inschriften ungeheuer; Kellermanns ernsthaftes und eifriges, leider zu früh unterbrochenes Beginnen sowohl wie das rasch unternommene und rascher aufgegebene pariser Projekt sind deutliche Zeichen des tief gefühlten Bedürfnisses.
关于编纂一部囊括全部拉丁铭文的总集之必要性，已毋庸赘证。众所周知，格鲁特–斯卡利杰 (Gruter-Scaliger) 之版本久已陈旧，慕拉托利 (Muratori) 之版本则从根本上既有缺陷又不周备；散布于两大总集之外、分布于上百部书之中的铭文数量惊人。凯勒曼 (Kellermann) 严肃认真、惜过早中止之尝试，与匆忙启动、又更匆忙放弃之巴黎项目，皆是深切需求之明证。
Wenn ich mir über die zweckmäßigste Art, demselben abzuhelfen, einige Worte erlaube, die zwar keinen Plan eines neuen Corpus Inscriptionum Latinarum, wohl aber die leitenden Ideen für einen solchen enthalten, so veranlaßt mich dazu ein zweijähriges, selten unterbrochenes Inschriftenstudium in dem Lande selbst, das immer Mittel- und Ausgangspunkt dieser Forschungen bleiben muß.
兹就解决此局之最适宜方式略陈数语；此非新《拉丁铭文大全》(Corpus Inscriptionum Latinarum) 之全盘计划，仅其指导性的诸般观念。促我言此者，乃我于该国本地几未间断之两年铭文研究，该国必为此类研究之手段与出发点。
[2]
Daß meine Andeutungen sich zunächst immer auf Italien beziehen, erklärt sich nicht minder aus der Natur der Sache als aus meiner persönlichen Stellung. Die Sammlung soll alle römischen Inschriften umfassen. Aber wo hören diese auf und wo fangen die mittelalterlichen an? Einigermaßen willkürlich ist jede Zeitgrenze; nicht unzweckmäßig indeß haben die Franzosen dafür das Ende des sechsten Jahrhunderts n. Chr. festgestellt, was sich besonders dadurch empfiehlt, daß die Konsulate, die vollständig zu geben wünschenswerth ist, um diese Zeit ganz aufhören.
我之提示始终先指意大利，此既因事物本身之性质，亦因我个人之处境。本总集当涵盖所有罗马铭文。然此范围何时终止、中世纪铭文何时开始？任何时间界限皆不免任意；然法国人将其定于公元六世纪之末，并非不合宜，此尤因，欲完整收录之执政官名表 (consulates) 恰于此时绝迹。
Zweck des C. I. L. ist, die sämmtlichen lateinischen Inschriften in eine Sammlung zu vereinigen, sie in bequemer Ordnung zusammenzustellen, dieselben nach Ausscheidung der falschen Steine in einem möglichst aus den letzten zugänglichen Quellen genommenen Text mit Angabe erheblicher varietas lectionis kritisch genau wiederzugeben und durch genaue Indices den Gebrauch derselben zu erleichtern. Kommentar ist wünschenswerth, nicht aber nothwendig.
《拉丁铭文大全》之宗旨在于：将全部拉丁铭文汇为一集；以便利之次序排列之；剔除伪刻 (falsche Steine) 之后，依尽可能取自最末端可达之来源所成之文本，附以重要异读 (varietas lectionis)，作批判性精确再现；并以精确之索引便利使用。评注 (Kommentar) 虽属可取，然非必需。
Editor's note · the four-clause programme
编者札记 · 四条纲领
This single sentence states Mommsen's entire programme: (1) collect all; (2) order conveniently; (3) sift forgeries and edit critically from the latest accessible sources, marking variants; (4) provide precise indices. The relegation of Kommentar to "desirable but not necessary" was a strategic choice, by giving up an exhaustive commentary, Mommsen made the project tractable. The four clauses still describe what every modern digital epigraphic corpus (EDH, EDR, EDCS, I.Sicily) does, with the addition that the database is the new "index."
此一句即蒙森全部纲领：（一）尽数收录；（二）以便利次序排列；（三）剔除伪刻、依最末端可达来源作批判性版本、标记异读；（四）提供精确索引。把"评注"贬为"可取而非必需"是一项战略选择，放弃穷尽性评注，蒙森使工程变得可行。这四条至今仍描述着任何现代数字铭文语料（EDH、EDR、EDCS、I.Sicily）所做之事，唯一新增者，是数据库本身已成为新的"索引"。
I. Sammlung des inschriftlichen Materials
I. Collection of Inscriptional Material第一部分 · 铭文材料之收集
Section gist本节大旨
Material comes from two sources: literature (printed corpora since Gruter, manuscript collections from the 15th–17th centuries, and the local-history books of Italian cities) and the stones themselves. Both must be re-examined: printed corpora must be checked against their manuscript sources; the stones must be inspected by autopsy because "all criticism without recourse to the ultimate sources is patchwork."
材料来自两端：一为文献（自格鲁特以降的印本总集、15–17 世纪的手抄本、意大利各城的地方志），二为石头本身。两者皆须重新审视：印本须比对其手抄本来源；石头须以亲眼勘察 (autopsy) 检验，因为"任何批判一旦不回溯至最末端之来源，便是拼凑之活。"
[3]
Das epigraphische Material ist theils aus der Literatur, theils aus den Steinen selbst zu entnehmen. In der Literatur sind das Wichtigste die großen epigraphischen Sammlungen, in denen sich im Wesentlichen schon vereinigt befindet, was bis zum J. 1770 von lateinischen Inschriften bekannt war.
铭文材料一部分自文献中取得，一部分自石头本身取得。文献中最重要者，是各大铭文总集，在其中实质上已汇集了至 1770 年止已知之拉丁铭文。
[4]
Indeß sind diese Sammlungen, deren man sieben bis acht rechnen kann, sämmtlich von der Art, daß man nicht bei ihnen stehen bleiben darf, sondern vielmehr ihren Quellen nachzugehen hat, deren Benutzung, wie sie von ihnen geschah, keinesweges den Forderungen unsrer Kritik entspricht. So z. B. hat Scaliger die Texte von Apian und Jacob Mazochi zum Theil recht willkürlich und ohne Angabe der überlieferten Lesart konstituirt. Viel weniger Verlaß ist nun erst auf Muratori, dessen Flüchtigkeit und Nachlässigkeit man Schritt vor Schritt kontroliren muß.
然这些总集，可数七八种，皆有此特点：不能止步于此，反须追溯其来源；这些来源被它们利用之方式，绝不符合我等批判之要求。例如，斯卡利杰 (Scaliger) 部分相当任意地构建阿皮安 (Apian) 与雅各布·马佐齐 (Jacob Mazochi) 之文本，未注明所传读法。对慕拉托利 (Muratori) 之信赖更须削减，其浮躁与疏忽，须一步一步加以核对。
[5]
Aber die Bücher reichen nicht aus; es muß, so viel es möglich ist, auf die Originale selbst zurückgegangen werden. Reist man doch nach Florenz und Paris, wenn man den Livius kritisch herausgeben will; wie sollte nicht der mit einer kritischen Ausgabe sämmtlicher Inschriften Beschäftigte die Einsicht der Steine selbst sich zu verschaffen suchen? Ja er muß es um so mehr thun, weil er sonst einen der Hauptvorzüge, den die Epigraphik vor der andern Literatur voraus hat, die unzweifelhafte, unanfechtbare Sicherheit des Textes, muthwillig aufopfern würde.
然书籍不足为凭；必须尽可能回到原物本身。欲批判性地校订李维 (Livius)，尚须远赴佛罗伦萨与巴黎；致力于全部铭文之批判性版本者，又岂可不亲眼察看石头？他必须如此而行，否则便擅自牺牲了铭文学相较于其他文学所独擅的主要优点之一：文本无可置疑、不容反驳之确定性。
[8]
Aber außer dem Studium dieser großen Sammlungen ist es für ein C. I. L. erforderlich, daß der Herausgeber desselben die Hauptfundörter der lateinischen Inschriften persönlich besuche. Nicht bloß werden dadurch fast immer ungedruckte, oft interessante Inschriften gewonnen, sondern, was noch wichtiger ist, es kommt eine gleichmäßige Behandlung in die ganze Sammlung und es wird fester Boden für die Kritik gewonnen, sowohl was die Aechtheit als was die Lesung betrifft.
然而，除研究这些大型总集之外，《拉丁铭文大全》尚必须由其编者亲身访问拉丁铭文之主要出土地。其益不仅在于由此几乎总能获得未刊行、且常常有趣之铭文；更重要者在于：全书可达均一处理之水准，且无论真伪还是读法，批判皆得坚实之基础。
Editor's note · the autopsy principle
编者札记 · 亲勘 (autopsy) 原则
The word Mommsen chose here is erforderlich, required, not wünschenswerth (desirable). With that single choice he turned autopsia from Borghesi's personal scholarly maxim into an institutional threshold for what would count as a critical edition. As Christer Bruun observes, this requirement "created an enormous amount of work for the CIL collaborators. No longer was it sufficient to cite a printed work from a previous century."
蒙森在此选用的词是 erforderlich ，必需的，而非 wünschenswerth（可取的）。仅此一词之选择，便把亲勘从 Borghesi 个人的学术格言，变为一部"批判性版本"得以成立的制度门槛。如 Christer Bruun 所言，此要求"为 CIL 之合作者群带来了庞大工作量。引用上一世纪某一印本，已不再充分"。
Bruun, "The Major Corpora," in Oxford Handbook of Roman Epigraphy (2015), p. 69.
[10]
Borghesi hat öfters gesagt: es sei ihm jede Collation einer alten Abschrift lieber als eine ganz neu gemachte Copie. Die Richtigkeit der Bemerkung ist einleuchtend; viel leichter macht man beim Abschreiben einen Fehler als eine falsche Berichtigung beim Collationiren, und bei bestrittenen, schwer lesbaren Inschriften ist die Entscheidung nur dann definitiv, wenn man die verschiedenen früheren Lesungen an dem Original selbst geprüft hat.
Borghesi 屡言：他宁取一旧抄之校勘 (collation)，不取一全新所作之抄本 (copy)。此言之确，一目了然，抄录时易生讹误，而校勘时反难误改；遇争议、难辨之铭文时，唯有就原物本身核对过各种早期读法，结论方得为定论。
II. Anordnung der Inschriften
II. Arrangement of the Inscriptions第二部分 · 铭文之排列
Section gist本节大旨
Two competing schemes exist: systematic (by content type, sacred, public, private) and topographic (by findspot, province, city). Mommsen rejects both as exclusive choices and proposes a hybrid: systematic for state-level material (decrees, magistrates, armies), topographic for everything else. His decisive argument: only a topographically organised corpus can absorb later discoveries without structural disruption.
既存两种相互竞争之方案：系统式（按内容分类，神事、公事、私事）与地形式（按出土地分类，行省、城邦）。蒙森否决"二者必择其一"之提法，主张一种混合方案：国家级材料（敕令、官员、军事）取系统式，其余取地形式。其决定性论点是：唯有地形式之总集，方能在不破坏结构的前提下吸纳后续发现。
[11]
Liegt nun die gesammte Masse des inschriftlichen Materials dem Arbeiter vor, so hat er zunächst dieselbe zu ordnen und die einzelnen Steine nach passenden Kategorien zusammenzustellen. Zwei Methoden sind in Eintheilung und Anordnung der römischen Inschriften bisher zur Anwendung gekommen: die eine legt den Inhalt der Inschriften zum Grunde und theilt dieselben zunächst in sacrale, öffentliche und private; die zweite geht aus vom Fundort und schließt sich an die geographische Eintheilung in Provinzen und städtische Distrikte.
当全部铭文材料置于工作者面前时，他首先须将之分类排序，把单块石头按合宜之范畴组合。罗马铭文之划分与排列，迄今所用者有两法：其一以铭文内容为本，先分为神事 (sacrale)、公事 (öffentliche)、私事 (private) 三类；其二以出土地为本，依省与城邑之地理划分。
Es scheint am gerathensten, keinem von beiden Systemen unbedingt zu folgen. Die römischen Inschriften haben nicht wie die griechischen nur die Sprache gemeinsam. Sie sind zum guten Theil Denkmäler eines Staats, und was diesen Staat betrifft, gehört zusammen, mag es auch an noch so fernen Punkten gefunden sein; so bilden die Inschriften einer Legion eine Einheit und sind keinesweges in all die verschiedenen Distrikte zu zersplittern, wo sie einmal kantonnirt hat.
最为妥当之策，似是对二者皆不无条件遵从。罗马铭文不同于希腊铭文，后者仅共享语言。罗马铭文中颇有部分是一个国家之纪念，凡涉此国家者属于一处，纵其于多远之地点发现亦然：故一支军团 (legion) 之铭文构成一整体，绝不应被分裂至该军团曾驻扎之各个分区。
[12]
Vor allem wichtig aber ist es, daß spätere Sammler nur auf ein nach topographischer Methode angelegtes Corpus mit Sicherheit fortbauen können. Den Lokalschriftstellern und Reisenden, nächst den Epigraphikern von Fach den wahren Konstituenten der Inschriftenkunde, blieb bei der bisherigen systematischen Ordnung der früher schon gesammelte Inschriftenvorrath des Ortes meistens unbekannt, weil er in den großen Sammlungen auseinandergerissen und nicht ohne viele Mühe wieder zu vereinigen war.
然最关键者：唯有按地形法编排之总集，后来者方能确信地继续向上构建。地方学者与旅行者乃铭文学之真正构成者（次于专业铭文学家）；在迄今所行之系统排列下，他们对其所在地早已收录之铭文素材，大多茫然无知，因为这些素材在大型总集中被拆散，不费大力气难以重新汇合。
Editor's note · cumulability as decisive criterion
编者札记 · 以"可累积性"为决定性标准
Mommsen's argument here is structural rather than aesthetic: only a geographically arranged corpus can absorb new finds without re-shuffling its existing entries. A content-classified corpus is a closed system; a topographic one is open-ended. This logic is the same logic that today drives every database keyed by inscription ID (EDH-ID, EDCS-ID, ISic-ID), each new inscription gets a number in its place without disturbing anyone else's numbers.
蒙森此处之论证是结构性的而非美学性的：唯有按地理编排之总集能在不重排既有条目的前提下吸纳新发现。按内容分类的总集是一个封闭系统；按地形分类者则是开放性的。此一逻辑与今日所有以铭文 ID 为主键的数据库（EDH-ID、EDCS-ID、ISic-ID）所遵循者完全相同，每一块新铭文获得一个属于其所在地的编号，而不扰动任何其他编号。
III. Kritik (Auszug)
III. Critical Editing (excerpt), text constitution, sigla, forgeries第三部分 · 文本批判（节选）， 文本构成、符号、伪刻
Section gist本节大旨
Mommsen distinguishes three editorial moves on every text: (a) text constitution from the best available source, (b) marking of editorial interventions, square brackets for losses restored by editor, round brackets for abbreviations expanded, angle brackets for surplus letters removed, and (c) judgment of authenticity. On forgery he is famously harsh: once an editor's deceitful intent has been proven, "probato dolo totum testem infirmari", his entire credibility is invalidated. Hundreds of items in Ligorio's 16th-century catalogue would later be condemned under this principle in CIL VI's inscriptiones falsae.
蒙森在每条文本上区分三种编辑动作：（甲）依最佳可达来源构建文本；（乙）标记编辑介入，以方括号表示编者补出之缺失字符、以圆括号表示编者展开之缩写、以尖括号表示编者删除之多余字符；（丙）真伪之判定。在伪刻问题上他著名地严厉：一旦编者之欺诈意图获证，"probato dolo totum testem infirmari"，其全部可信度即作废。利戈里奥 16 世纪目录中之数百条目，后来依此原则被列入 CIL VI 之"伪刻"(inscriptiones falsae) 部分。
Aufgabe der Kritik ist die Ausscheidung der falschen Inschriften und die Constituirung des Textes der echten. Beide Aufgaben sind in hohem Grade schwer, vor allem die erste. Daß man den Fälschungen eines Ligorius, eines Fourmont, eines Trigueros mißtraut, versteht sich; aber damit ist nur der erste Schritt gethan.
批判 (Kritik) 之任务是剔除伪刻、构建真刻之文本。二事皆甚艰难，首者尤甚。对利戈里奥 (Ligorius)、富尔蒙 (Fourmont)、特里盖罗斯 (Trigueros) 等人之伪造保持戒心，自不待言，然此仅是第一步。
V. Die Art der Ausführung
V. The Manner of Execution第五部分 · 实施之方式
Section gist本节大旨
Mommsen sketches budget, staff, and timeline. He demands the corpus be edited and printed in Germany, not Italy, on practical grounds (cheaper, faster, libraries lend books home) but with an implicit political subtext, the authoritative edition of Rome's epigraphic heritage will issue from Berlin. He estimates fifteen folio volumes of source material must be cut up just to assemble the working manuscript; approximately 80,000 stones must be reproduced. Above all he insists: better to postpone the project than to begin it with half-measures.
蒙森勾勒预算、人员与时间表。他主张《大全》之编辑与印行必须在德国、而非意大利，理由是实务上的（成本更低、速度更快、图书馆允许借书回家），然亦隐含政治背景：罗马铭文遗产之权威版本，将由柏林发出。他估算：仅为编出工作底稿，便须裁切十五卷对开本之源材料；约八万块石头须复制。其核心要求是：与其以半数资源、半数信任启动，宁可暂缓。
[27]
Das Unternehmen ist weitläufig und kostbar und verlangt großes Vertrauen zu dem damit Beauftragten. Sind bedeutende Geldkräfte und geeignete Individuen nicht zur Disposition, so ist es besser, dasselbe zu vertagen, als es mit halben Mitteln und halbem Zutrauen zu beginnen.
此事工程浩大、费用高昂，需对受托者投以重大信任。若无可观之财力与适当之人选可调配，则宁可暂缓，亦不可以半数资源、半数信任而启动。
[29]
Die Bearbeitung und Herausgabe des C. I. L. muß in Deutschland stattfinden, nicht in Italien. Hier sind klimatische und äußere Verhältnisse dem angestrengten und dauernden Arbeiten durchaus ungünstig und der Aufenthalt kostspieliger als in der Heimath.
《拉丁铭文大全》之编辑与出版必须在德国，而非意大利。此地之气候与外在环境对持续紧张之工作极为不利，且驻留之费用较本国为高。
[31]
Möchte aber endlich die Zeit gekommen sein, wo dies in Deutschland, Frankreich und Italien so oft begonnene, so ungeduldig erwartete Werk mit Ernst und Nachdruck begründet und seiner Vollendung zugeführt wird! Der gegenwärtige Augenblick bietet im Ganzen günstige Aussichten dar.
惟愿如下之时刻终于到来：此一在德、法、意三国数度肇始、被切盼之工程，能以严肃与坚定之态度奠基，并被引向完成。当下时刻，总体上呈现着有利之前景。
[32]
Rom, im Januar 1847.
于罗马，1847 年 1 月。
An interlude · on standardization, a question Mommsen did not answer in 1847
插话 · 关于标准化, 1847 年备忘录未答的一个问题
Before and after 1847: the slow consolidation of editorial sigla
1847 之前与之后, 编辑符号的缓慢统一
Read the 1847 memorandum again for a list of editorial signs. You will not find one. Mommsen specifies that variants should be "marked," that forgeries should be excised, that indices should be precise; he never specifies which brackets do what. The silence is consequential, because in 1847 the bracket conventions of European editing were genuinely incoherent, and they remained so for another 84 years. The section below traces this longer arc, drawing on the 1932 papers of Van Groningen and Wilcken (uploaded to the local mount), the local scholarship in documents/epidoc/inschriften, and lectures 6, 7, and 8 of the course this matrix supports.
把 1847 年备忘录翻回来看一遍, 找一份"编辑符号清单"。你找不到。蒙森的确说要"标记"异读、要剔除伪刻、索引要精确, 但他从不指明用哪些括号表示哪些情况。这一沉默并非无足轻重, 因为 1847 年欧洲编辑界的括号实践确实是不统一的, 而这一状态又持续了 84 年。下文追溯这条更长的弧线, 所据为本地挂载之 1932 年 Van Groningen 与 Wilcken 二位之文章、documents/epidoc/inschriften 下之本地学术资料, 以及本矩阵所支持课程之第 6、7、8 讲。
§A Before 1847: every editor invented their own sigla1847 之前: 每个编辑者都自创一套符号
The systematic use of marks to flag editorial interventions begins, for the printed era, with Erasmus's Novum Instrumentum Omne (Basel, 1516). Erasmus used italics in the Vulgate column to flag where his Greek-based correction differed from the received Latin reading; the practice was an editorial signal aimed at the reader, not yet a sigla system. Across the 16th and 17th centuries, the practice diversified. Italian humanist editors of inscriptions (Smetius 1588, Gruter 1602) developed thematic and topographic indices but not a uniform bracket convention. Eighteenth-century editors continued to invent their own marks: Maffei's Prospectus Universalis Collectionis (1732) proposed a unified corpus but did not propose a unified set of signs.
编辑介入用符号系统化标示, 在印本时代肇始于伊拉斯谟的《新约全书》(Novum Instrumentum Omne, Basel, 1516)。伊拉斯谟在武加大 (Vulgate) 译文一栏中用斜体标示自己依希腊文所作修正与流行拉丁读法不同之处, 这一做法是面向读者的编辑信号, 还不是一套符号体系。16 至 17 世纪间, 实践多元化。意大利人文学派之铭文编者 (Smetius 1588, Gruter 1602) 发展出主题与地形之索引, 却未发展出统一的括号惯例。18 世纪的编者继续自创各自的符号: Maffei 的《合一总集预案》(Prospectus Universalis Collectionis, 1732) 提议建立统一的语料库, 却未提议统一的符号体系。
A first systematic attempt appears in August Boeckh's preface to CIG I (1828), §VI: the "diplomatic principle." Boeckh's formulation is lapidary in Latin: nullam in eum immisimus coniecturam, "we have inserted no conjecture into it." All conjectural supplements were to be marked with square brackets, variant readings noted in the apparatus, prior restorations indicated below but never allowed to enter the transcription. This is the clearest pre-Leiden methodological declaration. But it applies to Greek inscriptions, and Hermann's school immediately contested whether Boeckh's principles even held for textual criticism in general. The Hermann–Boeckh controversy (1825 onward) shows that there was no European consensus on what brackets meant.
第一次系统性尝试见于勃克 CIG I (1828) 序言第六节, 即"原文原貌原则"(diplomatic principle)。勃克之拉丁表述简洁如石: nullam in eum immisimus coniecturam, "我等未在文本中放入任何臆改"。一切臆补皆须以方括号标出, 异读记于校勘附录, 前人之复原置于正文之下, 但绝不进入抄文本身。这是莱顿规约之前最清晰的方法论宣言。但其适用对象是希腊铭文, 且赫尔曼 (Hermann) 学派立即提出抗议, 质疑勃克之原则能否用于一般文本批判。1825 年起之赫尔曼–勃克之争表明: 欧洲学界对"括号意味着什么"并无共识。
§B Mommsen's silence in 1847蒙森在 1847 年的沉默
Search the memorandum for a sigla specification. Section III ("Kritik") demands that forgeries be excised and that variant readings be marked. Section II argues for topographic over systematic arrangement. Section IV proposes the index architecture. Nowhere is there a list of which bracket means which editorial intervention. Mommsen the rigorous editor, who specified that fifteen folio volumes must be cut up for the working manuscript and that ~80,000 stones must be reproduced, assumed the basic editorial toolkit was inheritable from existing print custom. He did not pause to specify it.
把备忘录翻一遍, 找符号清单。第三节 ("Kritik", 文本批判) 要求剔除伪刻、标记异读。第二节论证地形式优于系统式。第四节提出索引架构。但全文之中并无一份"哪种括号表示哪种编辑介入"之清单。蒙森其人, 那位精细到要求"十五卷对开本之源材料须裁切以制工作底稿、约八万块石头须复制"的编者, 在符号问题上却默认: 既有印本传统所传之基本编辑工具箱足以使用, 无须停下来加以规定。
This was a structural omission, not an oversight. Mommsen was building an institutional Großbetrieb; he assumed that each editor in the network would bring print-shop literacy with them. But "print-shop literacy" in 1847 was not one thing. It was the residue of three centuries of accumulated, inconsistent, sometimes contradictory editorial practice. 真正的问题恐怕是: a corpus that is internally consistent in coverage (collect all), in arrangement (topographic), and in source-criticism (autopsy required) can still be internally inconsistent in its sigla. And that is what happened to the CIL.
这并非疏忽, 而是结构性的省略。蒙森是在建造一座大企业 (Großbetrieb), 他默认网络中每一位编辑都会自带印刷行业的基本素养。但 1847 年的"印刷行业素养"并不是单一的东西, 而是三百年间累积、不一致、有时相互矛盾的编辑实践之沉淀。真正的问题恐怕是: 一部语料在覆盖面 (尽数收录)、排列法 (按地理) 与来源批判 (亲勘必须) 上都可以保持内部一致, 但在符号上仍可以内部不一致。这, 正是 CIL 的实际遭遇。
§C The cost of the silence: CIL's internal heterogeneity, 1853–1899沉默的代价: CIL 自身的不一致, 1853–1899
Open any two CIL volumes from different decades and the bracket conventions will not align perfectly. The Italian volumes Mommsen edited himself (IX–X, 1883) lean toward Borghesi-school usage. The German volumes (II Spain by Hübner, 1869; III Danubian by Mommsen and others, 1873) shift conventions across editors. The German-school habit of double angle brackets « » for editorial expansion competes with the French-school habit of round brackets ( ); the Italian habit of dotting uncertain letters competes with the German habit of italicizing them. Inside the CIL there is no enforced standard; inside the corresponding Greek IG (which begins to appear in 1903) the same story unfolds. As Wilcken would later observe in 1932, this "geradezu unerträglichen Diskrepanz" ("downright unbearable inconsistency") had become the rule.
把两卷不同年代之 CIL 摆在一起看, 括号惯例不会完全对齐。蒙森自己编的意大利诸卷 (IX–X, 1883) 倾向博尔盖西学派之用法; 德国学者所编诸卷 (II 西班牙, Hübner 1869; III 多瑙诸省, 蒙森等 1873) 在不同编者之间又有切换。德国学派常用双尖括号 « » 表示编者补字, 法国学派则用圆括号 ( ); 意大利学派用字下加点表示读法存疑, 德国学派则用斜体。CIL 内部无强制标准; 相应之希腊《铭文集成》(IG, 自 1903 年起出版) 亦演同一剧本。Wilcken 后来在 1932 年所言之 "geradezu unerträglichen Diskrepanz" ("近乎无法忍受的不一致"), 已成为常态。
The practical cost was tangible. As the user's course lecture on brackets puts it: "在同一页书里, 不同的地方, 有时引的是作者文本, 有时引的是取自纸草的文书, 有时又引的是碑刻残片; 而这些地方的不确定现象, 由于表达方式不同, 在读者眼前看起来也就不同" (Wilcken 1932, in Chinese paraphrase from lecture 7). One scholar's [α] and another scholar's α̣ might refer to the same uncertainty about the same letter on the same stone, but the print reader could not tell.
实际代价是可见的。如本课程"括号"一讲所述: 同一页书里, 不同处, 有时引作者文本, 有时引纸草, 有时引碑刻; 而这些地方"由于表达方式不同, 在读者眼前看起来也就不同" (Wilcken 1932, 第 7 讲中文转述)。某位学者写 [α], 另一位写 α̣, 可能指的是同一块石头同一个字母的同一种不确定, 但印本读者无从辨别。
§D 1931: Leiden, and the convention that filled the gap1931 年: 莱顿, 与填补这一空缺的规约
The institutional initiative came from outside epigraphy. The Union Académique Internationale, founded after the First World War, commissioned the Belgian scholar J. Bidez (Ghent) and the Dane A. Drachmann to draft a brochure on the use of critical signs in scholarly editions. Their preliminary draft (Avant-projet d'une brochure traitant de l'emploi des signes critiques et de la rédaction de l'apparat critique dans les éditions savantes) circulated among papyrologists, who then took up the question at the 18th International Orientalist Congress, held at Leiden from 7 to 12 September 1931.
制度性的推动来自铭文学之外。第一次世界大战后成立的国际学术联盟 (Union Académique Internationale) 委托比利时根特学者 J. Bidez 与丹麦学者 A. Drachmann 起草一本关于"学术版本中如何使用校勘符号"的小册子 (brochure)。其预备草案 (Avant-projet d'une brochure traitant de l'emploi des signes critiques et de la rédaction de l'apparat critique dans les éditions savantes) 在纸草学界先行传阅, 而后被提交给 1931 年 9 月 7 至 12 日于莱顿召开的第十八届国际东方学家大会。
The Papyrology section of that congress, after hearing reports by Bell, Hunt, and Van Groningen, and comments by Schubart and Vitelli, charged a three-person drafting committee: Jouguet (Paris), Van Groningen (Groningen), and Hombert (Brussels). They produced a small system, eleven signs in all, and the section published it in two simultaneous places: in Latin in Van Groningen's De signis criticis in edendo adhibendis in Mnemosyne 59 (1932), and in French in his Essai d'unification des méthodes employées dans les éditions de papyrus in Chronique d'Égypte VII no. 13–14 (January 1932), pages 285 ff. Both papers are now in the local mount.
大会纸草学分会听取了 Bell、Hunt、Van Groningen 三位报告, 听取 Schubart 与 Vitelli 二位意见后, 委托三人起草委员会: Jouguet (巴黎)、Van Groningen (格罗宁根)、Hombert (布鲁塞尔)。他们拟出一套小型符号体系, 共十一种, 并由分会同时发表于两处: 拉丁语者为 Van Groningen 的《关于在校勘中所用之符号》(De signis criticis in edendo adhibendis), 刊于《Mnemosyne》59 (1932); 法语者为同作者之《纸草版本中所用方法统一之试拟》(Essai d'unification des méthodes employées dans les éditions de papyrus), 刊于《Chronique d'Égypte》第 7 卷第 13–14 号 (1932 年 1 月), 第 285 页起。两份文献现皆在本地挂载之内。
§E The agreed sigla, in five lines议定之符号, 五行
The system of 1931–1932, lightly revised in 1982, is still in use. The core:
1931 至 1932 年议定、1982 年小幅修订之体系, 沿用至今。其核心:
- [ ] square brackets: editor restores text lost from the stone方括号: 编者补全石上已失的文字
- ⟨ ⟩ angle brackets: editor supplies letters that were never on the stone (scribe's omission)尖括号: 编者补入石上本无之字母 (书写者漏刻)
- ( ) round brackets: editor expands an abbreviation present on the stone圆括号: 编者展开石上确有之缩写
- { } curly brackets: editor removes surplus letters that are on the stone but should not be in the text花括号: 编者删去石上虽有、但不应在文本中的多余字母
- ⟦ ⟧ double square brackets: erasure of text once on the stone, often by damnatio memoriae双方括号: 石上原有、后被擦除之文字, 常为除忆诅咒 (damnatio memoriae) 所致
- α̣ under-dot: uncertain reading (the letter is on the stone but ambiguous)字下加点: 读法存疑 (字母在石上但模糊)
Wilcken's German announcement in Archiv für Papyrusforschung 10 (1932), pages 211–212 added the institutional weight: as the editor of Inscriptiones Graecae after Wilamowitz's death (1931), he committed IG to the new sigla, and urged all epigraphic publishers, Latin as well as Greek, to follow. The famous closing wish: "Es ist unsere Hoffnung und unser Wunsch, daß dies Beispiel überall in der Epigraphik Nachahmung finden möge, natürlich in der lateinischen ebenso wie in der griechischen."
Wilcken 在《Archiv für Papyrusforschung》10 (1932), 211–212 页之德文公告添加了制度性分量: 作为 Wilamowitz 1931 年逝世后接手主持《Inscriptiones Graecae》的编辑, 他承诺 IG 此后采用新符号, 并力促所有铭文出版者 (无论拉丁与希腊) 一并跟从。著名的结句之愿: "Es ist unsere Hoffnung und unser Wunsch, daß dies Beispiel überall in der Epigraphik Nachahmung finden möge, natürlich in der lateinischen ebenso wie in der griechischen" (我们的希望与愿望是, 此例能在全部铭文学中获得仿效, 拉丁与希腊一并如此)。
§F After 1932: rasura, Dow 1969, and EpiDoc's digital codification1932 年之后: 擦除问题、Dow 1969、与 EpiDoc 之数字编码
One area resisted clean codification: rasura, the deliberate erasure of names from public inscriptions. The Leiden brackets cover the static case (text once on the stone, now erased, marked with ⟦ ⟧), but real epigraphic erasures are layered. A senator's name is chiselled out under Domitian; another name is later cut over the empty space; both layers are part of the historical record. As Peter Rhodes (2018) documents and the user's course lecture 8 on rasura discusses at length, the Soviet Great Soviet Encyclopaedia's 1953 erasure of Beria is the modern epigraphic counterpart: erasure as a public act that leaves a visible wound. Leiden's ⟦ ⟧ flags the wound but cannot fully encode the layering.
有一处难以干净编码: rasura, 即公共铭文上对人名的有意擦除。莱顿括号能处理静态情形 (文字曾在石上、现被擦除, 以 ⟦ ⟧ 标示), 但真实的铭文擦除是有层次的。某元老的名字在多米提安治下被铲去, 后来另一名字又刻在原空白之上; 这两层都是历史记录的一部分。如 Peter Rhodes (2018) 所记、本课程关于 rasura 的第 8 讲所详论: 1953 年苏联《伟大苏联百科全书》对贝利亚 (Beria) 的擦除, 正是现代版的铭文式擦除, 一种公共行动, 留下可见之伤口。莱顿之 ⟦ ⟧ 能标示伤口, 却无法完整编码层次。
In 1969 Sterling Dow proposed a "suggested reformulation" of the Leiden system in his Conventions in Editing, tightening some ambiguous cases. In 1982 the conventions were lightly revised. From 2000 onward, the EpiDoc community (Tom Elliott, Gabriel Bodard, Hugh Cayless, Charlotte Roueché and others) encoded the entire Leiden vocabulary as TEI/XML elements: <supplied reason="lost"> for square brackets, <supplied reason="omitted"> for angle brackets, <expan><abbr/><ex/></expan> for round brackets, <surplus> for curly brackets, <del rend="erasure"> for the rasura case, <unclear> for the under-dot. The Leiden brackets, born in a Dutch hotel room in September 1931, now survive as machine-readable XML.
1969 年, Sterling Dow 在其《Conventions in Editing》中提出莱顿体系的"建议性重述", 收紧若干含混情形。1982 年, 规约获小幅修订。自 2000 年起, EpiDoc 社群 (Tom Elliott、Gabriel Bodard、Hugh Cayless、Charlotte Roueché 等) 把整套莱顿词汇编码为 TEI/XML 元素: <supplied reason="lost"> 对应方括号, <supplied reason="omitted"> 对应尖括号, <expan><abbr/><ex/></expan> 对应圆括号, <surplus> 对应花括号, <del rend="erasure"> 对应擦除, <unclear> 对应字下点。这套 1931 年 9 月在一间荷兰旅馆房间内议定的莱顿括号, 如今以机器可读的 XML 形式存活下来。
The loop closes. Mommsen 1847 did not specify sigla; the CIL inherited the resulting inconsistency for two generations; the 1931 Leiden meeting produced the missing standard; EpiDoc made that standard machine-checkable. The point here is not that Mommsen failed to anticipate this — no 1847 plan, however prescient, could have specified everything its successors would need. Standardizing editorial sigla turned out to be a separate problem, taking a separate institutional convening 84 years later. The open question is which other assumptions the 1847 plan made implicitly but did not specify, and which of those will need their own "1931 Leiden moment" between now and 2125.
一个完整的回环。蒙森 1847 年没有规定符号; CIL 在两代人间承袭了这一不一致; 1931 年莱顿会议补上了所缺的标准; EpiDoc 让标准变得机器可校验。这并非蒙森失察，任何 1847 年的纲领, 不论多么有远见, 都不可能事先指定其后继者所需的一切。编辑符号的标准化是一个独立的问题, 由独立的制度性会议解决, 在 1847 年之后 84 年。剩下的开放问题是: 1847 年的纲领还隐含地假定了什么、却未明文规定? 哪些会在今日至 2125 年之间, 需要属于它们自己的"1931 莱顿时刻"?
Sources consulted for this section
本节所参考之资料
- Wilcken, Ulrich. "Das Leydener Klammersystem." Archiv für Papyrusforschung 10 (1932), 211–212. [local upload:
1932 wilcken das leydener klammersystem copy-41457f20.pdf]
- Van Groningen, B.A. "De signis criticis in edendo adhibendis." Mnemosyne 59 (1932), 362–365. [local upload:
1932 van groningen de signis criticis in edendo adhibendis.pdf]
- Van Groningen, B.A. "Essai d'unification des méthodes employées dans les éditions de papyrus." Chronique d'Égypte VII no. 13–14 (1932), 285 ff. [local upload:
1932 von groningen essai d unification des méthodes employées dans les éditions de papyrus.pdf]
- Dow, Sterling. Conventions in Editing: A Suggested Reformulation of the Leiden System. Durham, NC: Duke University, 1969.
- Rhodes, P.J. "Erasures in Greek Public Documents." In The Materiality of Text, ed. Petrovic, Petrovic & Thomas. Leiden: Brill, 2018, 145–166.
- Bodard, Gabriel. "EpiDoc: Epigraphic Documents in XML for Publication and Interchange." In Latin on Stone, ed. F. Feraudi-Gruénais. Lanham: Lexington, 2010, 101–118.
- Course lectures, this matrix supports:
20260408 intro epigraphy 6 brackets.pdf; 20260415 intro epigraphy 7 leiden.pdf; 20260422 intro epigraphy 8 rasura.pdf.
- Local mining reports:
symbolic-system-revised.md (esp. §V "Standardization and the Digital Era 1931–Present"); chapter4-english-revised.md §5.
After 1847 · how the project Mommsen proposed actually unfolded
1847 之后 · 蒙森所提工程实际如何展开
From the Berlin memorandum to the digital corpus
从柏林备忘录到数字语料库
The five sections above set out a programme; this section sketches its reception. Drawing on local scholarship mounted under documents/epidoc/inschriften/scholarship, the following nine subsections track how each of Mommsen's principles fared between 1847 and 2025, including those that survived, those that broke, and the new problems that 21st-century digital corpora face. Citations point to the specific .md files and book chapters consulted.
上文五节铺陈的是一个纲领；本节勾勒其受容史。所据为挂载于 documents/epidoc/inschriften/scholarship 之本地学术资料。下方九小节追踪蒙森各项原则在 1847 至 2025 年之间的命运，含其得存者、其断裂者、以及 21 世纪数字语料所面临的新问题。引文指向所参考的具体 .md 文件与书章节。
§1 From manifesto to printing press: CIL 1853–1899从纲领到印行：CIL 1853–1899
Six years after Mommsen's memorandum, in 1853, the Berlin Academy approved the CIL plan and Mommsen began work in earnest. The first volume (CIL I, Inscriptiones Latinae antiquissimae ad C. Caesaris mortem) appeared in 1863. Subsequent volumes followed at irregular intervals over half a century. The Italian volumes (CIL IX–X, edited by Mommsen himself) appeared in 1883, the volume that contains, on its page corresponding to Panhormus / Palermo, entry CIL X 7296: the small bilingual stonecutter's sign that recurs throughout the rest of this matrix. Mommsen's 1847 dictum "es ist besser, dasselbe zu vertagen, als es mit halben Mitteln zu beginnen" governed the timeline: when funding wavered or collaborators failed, work paused rather than degraded.
蒙森呈交备忘录六年之后，1853 年柏林学院批准了 CIL 之计划，蒙森始作实质性工作。第一卷（CIL I，《最古至凯撒之死的拉丁铭文》）于 1863 年出版。其后各卷以不等之间隔陆续问世，前后逾半世纪。意大利部分（CIL IX–X，由蒙森本人编辑）在 1883 年出版，其中相应 Panhormus / 巴勒莫 (Palermo) 一页所载即为 CIL X 7296：贯穿本矩阵其余各页之那块双语小型石匠招牌。蒙森 1847 年那句"宁可暂缓，亦不可以半数资源启动"主宰了进度：每当资金动摇或合作者掉队，工程便暂停，而非降低质量。
Between 1853 and 1903 the CIL absorbed approximately 180,000 inscriptions across seventeen volumes. The collaborator network ran from Italy to Britain, North Africa, the Balkans, and the eastern provinces. Christer Bruun summarises: "In the history of Graeco-Roman epigraphy, this son of a Lutheran minister from the border region of Schleswig … towers above all other scholars. Mommsen was a 'militant epigrapher' in the highest degree, but he was also … a marvellous organiser of the world-wide scholarly community." Bruun 2015, 68
1853 至 1903 年间，CIL 在十七卷中吸纳约 180,000 条铭文，合作网从意大利伸至英国、北非、巴尔干与东方诸省。Christer Bruun 总结此一制度成就："在希腊–罗马铭文学史中，这位来自石勒苏益格 (Schleswig) 边境地区一位路德宗牧师之子，凌驾于其他所有学者之上。蒙森是最高程度上一位'好战的'铭文学者，亦是一位卓越的世界性学术社群组织者。"Bruun 2015, 68
§2 The Großbetrieb der Wissenschaften, the industrial humanities学术之"大工业经营"， 人文学之工业化
Adolf Harnack, Mommsen's contemporary at the Berlin Academy, characterised the CIL's mode of production with a now-famous phrase: Großbetrieb der Wissenschaften, "large-scale production of the sciences." Chad Wellmon has shown that Mommsen self-consciously imported into humanistic scholarship the management techniques that contemporaries were applying to industrial labour: division of tasks, standardised procedures, centralised supervision, parallel field work and editorial work, and a hierarchical network of Mitarbeiter. The Linnaean analogy is exact: "Like Linnaeus, who oversaw a global network of naturalists sending specimens to Uppsala, Mommsen managed a network of collaborators." Wellmon 2019
蒙森在柏林学院之同代人哈尔纳克 (Adolf Harnack) 以一个如今著名的术语描述 CIL 之生产方式：学术之大工业经营 (Großbetrieb der Wissenschaften)。Chad Wellmon 已表明，蒙森自觉地把同时代工业劳动所采用的管理技术输入到人文学术之中：分工、标准化程序、集中监督、田野与编辑工作并行、由 Mitarbeiter（合作者）构成的层级网络。林奈式的类比精确无误："正如林奈监管一个把标本送到乌普萨拉的全球博物学家网络，蒙森管理一个合作者网络。"Wellmon 2019
This industrial logic is not incidental; it is what makes Mommsen's 1847 memorandum a foundational text not merely for epigraphy but for the modern model of "big humanities", large, multi-decade, state-funded, hierarchically organised scholarly projects whose product is an indispensable infrastructure rather than a single monograph. Every subsequent digital corpus (EDH, EDCS, EDR, I.Sicily, EpiDoc) inherits this organisational pattern. chapter4-english-revised.md §3
此一工业化逻辑非偶然，它正是 1847 年备忘录之所以不仅作为铭文学、更作为现代"大人文"模型之奠基性文献的原因，大型、跨越数十年、由国家资助、层级化组织、其产品乃不可或缺之基础设施而非单一专著的学术工程。其后所有数字语料（EDH、EDCS、EDR、I.Sicily、EpiDoc）都继承这一组织模式。chapter4-english-revised.md §3
§3 Autopsy in practice, and the educated eye亲勘 (autopsy) 的实践 ， 并须以受过训练的眼睛
Mommsen's erforderlich ("required") elevated personal inspection of the stones from Borghesi's habit to an institutional rule. But as Emil Hübner, Mommsen's most important Iberian collaborator, wrote already in 1870, autopsy is never naive: "das Inschriftenlesen ist vielmehr eine Kunst, die wie alle Künste und Fertigkeiten technische Vorkenntnisse und dauernde Uebung voraussetzt" [reading inscriptions is rather an art which, like all arts and skills, presupposes technical knowledge and constant practice]. Hübner 1870; epigraphic-method-revised.md §1
蒙森的"必需" (erforderlich) 一词把对石头的亲眼勘察从 Borghesi 的个人习惯上升为制度规则。但正如蒙森在伊比利亚最重要的合作者 Emil Hübner 于 1870 年所言，亲勘从不是天真的："读铭文毋宁是一门艺术，如同所有艺术与技艺一样，它要求事先具备技术知识并须持续操练。"Hübner 1870；epigraphic-method-revised.md §1
Marco Buonocore reinforces this from the manuscript side: when planning the CIL, Mommsen "realized that to achieve a level of accuracy beyond that of the existing printed collections of inscriptions, it would be necessary to take account of the entire manuscript tradition which, from the Carolingian age down to the nineteenth century, had collected and preserved important information about epigraphic texts." Buonocore 2015, 23 Autopsy did not replace the manuscript tradition, it disciplined the use of that tradition by giving the editor a physical referent against which copies could be measured.
Marco Buonocore 自手稿一侧补充此点：蒙森在规划 CIL 时"意识到，要达到超越既有铭文印本之精确度，必须把从加洛林时代到 19 世纪之间收集和保存过铭文重要信息的整个手稿传统纳入考量"。Buonocore 2015, 23 亲勘并未取代手稿传统，它规训了对此传统之使用，藉给编者一个可与抄本相对照的物质参照。
§4 The forensic purge, Ligorio and probato dolo司法式清查 ， 利戈里奥与"probato dolo"原则
Section III of the memorandum demanded that forgeries be excised. The CIL editors put this into practice with a famous principle Mommsen formulated in CIL X p. x: probato dolo totum testem infirmari, "once his deceitful intent has been proven, his entire credibility as a source is invalidated." Hundreds of inscriptions known only through Ligorio's forty-volume manuscript catalogue were marked false in the inscriptiones falsae sections of CIL: CIL VI alone contains 3,643 such items. Orlandi et al. 2015, 43–45
备忘录第三节要求剔除伪刻。CIL 编辑实际执行此要求时，依蒙森在 CIL X 第 x 页所拟之著名原则：probato dolo totum testem infirmari，"一旦其欺诈意图获证，其整体作为来源之可信度便作废。"仅经由利戈里奥四十卷手稿目录而知的数百条铭文在 CIL 的"伪刻" (inscriptiones falsae) 部分被标为伪，仅 CIL VI 一卷便含 3,643 项。Orlandi et al. 2015, 43–45
Modern scholarship has substantially complicated Mommsen's binary purge. Orlandi, Caldelli and Gregori have shown that Ligorio was "moved by the desire to 'give the dead their souls back' (restituite l'anima agli estinti)," and roughly one-fifth of the seventy stone-carved forgeries originally condemned under probato dolo "have now been rehabilitated as genuine." Orlandi et al. 2015, 45, 50 John North Hopkins and Sarah McGill have pushed further: "reading is equated with policing" in Mommsen's framework, and the forensic optic obscures what forgeries reveal as historical texts and creative works in their own right. Hopkins & McGill 2023
现代学术大幅复杂化了蒙森的二元清查。Orlandi、Caldelli 与 Gregori 已表明，利戈里奥"由'让死者重得灵魂' (restituite l'anima agli estinti) 之愿望所驱动"，而原本依 probato dolo 被判伪的七十块石刻伪造中，约五分之一"已被恢复为真品"。Orlandi et al. 2015, 45, 50 Hopkins 与 McGill 推得更远：在蒙森的框架下"阅读被等同于警务执法"，而这套司法式眼光遮蔽了伪刻作为自身的历史文本与创造性作品所揭示之内容。Hopkins & McGill 2023
§5 Topographic arrangement settles, and becomes the digital primary key地形式排列被定为基准 ， 并成为数字主键
Section II's hybrid systematic / topographic scheme did not survive intact. In practice, the CIL volumes are organised almost entirely by region: vol. II Iberia, vol. III the Danubian and eastern provinces, vol. V northern Italy, vol. VI Rome itself, vols. IX–X southern Italy and Sicily, vol. XIII the German provinces and Gaul, vol. XV the urban instrumentum. Mommsen's argument from cumulability won decisively. Every modern digital corpus uses the same logic, with the city / province / findspot as its primary axis of organisation: I.Sicily orders inscriptions by Sicilian city; EDH classifies by province; even the cross-database identifiers (EDR-N, EDCS-N) presuppose that a stone has a definite findspot.
第二节那种"系统/地形"混合方案并未完整存活。实际操作中，CIL 各卷几乎完全按地区编排：第 II 卷伊比利亚、第 III 卷多瑙诸省与东方诸省、第 V 卷意大利北部、第 VI 卷罗马本城、第 IX–X 卷意大利南部与西西里、第 XIII 卷日耳曼与高卢、第 XV 卷都城的器物铭。蒙森基于"可累积性"的论证决定性地胜出。所有现代数字语料以同一逻辑运作，皆以城市 / 行省 / 出土地为组织主轴：I.Sicily 按西西里城市排序；EDH 按行省分类；甚至跨数据库的标识符 (EDR-N、EDCS-N) 也预设每块石头有确定的出土地。
§6 The Leiden Conventions of 1931, international standardisation Mommsen anticipated1931 年莱顿规约 ， 蒙森所预见的国际标准化
Mommsen's Section III tacitly assumed a system of sigla, square, round, angle, curly brackets, but no two CIL volumes are entirely consistent in their use. The 1931 Leiden Congress (papyrology section, on the initiative of the Union Académique Internationale) fixed this by negotiating a single international system. B.A. Van Groningen's foundational article in Mnemosyne 59 (1932) stated the universalist principle in Latin: "necessarium esse omnes, si fieri possit, posthac in omnibus semper idem agere" [it is necessary that everyone should henceforth, if possible, always do the same thing in all cases]. Van Groningen 1932, 362; symbolic-system-revised.md §V
蒙森第三节默认了一套符号系统，方括号、圆括号、尖括号、花括号，然实际上 CIL 各卷之使用并不全然一致。1931 年莱顿大会（纸草学组，由国际学院联盟倡议）以协商出一套统一的国际系统解决了此问题。B.A. Van Groningen 在《Mnemosyne》59 (1932) 之奠基性文章以拉丁语陈述其普遍主义原则："necessarium esse omnes, si fieri possit, posthac in omnibus semper idem agere"，"必要的是：自今而后所有人，凡可能之时，于一切情况下皆做同一件事。"Van Groningen 1932, 362；symbolic-system-revised.md §V
The Conventions specify twelve canonical signs: square brackets for restored lacunae, angle brackets for editorial supplements, round brackets for abbreviation expansions, curly brackets for surplus, underdots for uncertain readings. Sterling Dow put the pragmatic case bluntly: the system is "a style sheet on which a large majority of persons have agreed … so that no one will be penalised for following it." Dow 1952, 115 What Mommsen had hoped for in 1847, a transnational consensus on critical practice, was achieved 84 years later, in a sphere that crossed papyrology and epigraphy alike.
莱顿规约规定十二个标准符号：方括号表示编者补出之阙文；尖括号表示编者补入；圆括号表示编者展开之缩写；花括号表示删去多余字符；字下加点表示读法存疑。Sterling Dow 把实用面说得直白：此系统是"一份大多数人已达成共识之样式表 …… 任何人按此行事皆不会受罚"。Dow 1952, 115 蒙森 1847 年所希望的，在批判实践上之跨国共识，在 84 年后达成，且范围横跨纸草学与铭文学。
§7 Digital corpora, EDH, EDR, EDCS, I.Sicily, EAGLE数字语料 ， EDH、EDR、EDCS、I.Sicily、EAGLE
From the 1990s, three Latin databases independently digitised the CIL's textual material, each with a different editorial philosophy. Manfred Clauss's EDCS at Frankfurt prioritised coverage: today over half a million inscriptions, but with limited tools for tracking which edition any given text came from. Géza Alföldy's EDH at Heidelberg prioritised criticism: ~80,000 inscriptions, each accompanied by an apparatus and a confidence rating. Silvio Panciera's EDR at Rome focused on the Italian peninsula and on linking each entry to a photograph. Together they constitute the digital descendants of CIL, the "letzten Quellen" (latest sources) Mommsen's editor was required to consult. Bodel/Prag/Roueché 2024; chapter4-english-revised.md §5
自 1990 年代起，三个拉丁数据库独立地把 CIL 之文本材料数字化，各持不同编辑哲学。Manfred Clauss 在法兰克福主持的 EDCS 以覆盖率为先：今日逾五十万条铭文，然其追溯各条文本所属版本之工具有限。Géza Alföldy 在海德堡主持的 EDH 以批判为先：约八万条铭文，每条皆附校记与可信度评级。Silvio Panciera 在罗马主持的 EDR 聚焦意大利半岛，并致力于让每一条目挂接照片。三者共同构成 CIL 之数字后裔，蒙森编者所必须查阅的"letzten Quellen"（最末端来源）。Bodel/Prag/Roueché 2024；chapter4-english-revised.md §5
Yet the interoperability problem is severe. As Kaše, Heřmánková and Sobotkova show, "in the case of other attributes, like the type of inscription, the mapping is much less straightforward as the two databases use different classification schemes", EDCS and EDH disagree on how to classify the same stone. The SDAM ETL pipeline from Aarhus University was built specifically to harmonise EDR + EDH + EDCS into a single linked dataset. Kaše et al. 2022; see sdam-paper-jdh2021.html
然其互操作问题严峻。Kaše、Heřmánková 与 Sobotkova 已表明，"在其他属性如铭文类型上，映射远非如此直白，两个数据库采用不同的分类方案"，EDCS 与 EDH 对同一块石头之分类常不一致。Aarhus 大学的 SDAM ETL 流水线正是为把 EDR + EDH + EDCS 协调为单一互连数据集而构建。Kaše et al. 2022；参 sdam-paper-jdh2021.html
§8 EpiDoc and FAIR, encoding the editorial decisions, not just the textEpiDoc 与 FAIR ， 所编码者非仅文本，而是编辑判断
Mommsen's Section III implicitly required that an edition record three things on every line: what is on the stone, what the editor restores, and how certain the editor is. Print could only carry two of these at a time. EpiDoc, the TEI-XML customisation in use across the EpiDoc-Aphrodisias project (2001 onward), I.Sicily, Telamon, IRT, IRCyr and others, encodes all three. <supplied reason="lost">, <gap reason="lost">, <unclear>, <choice><reg/><orig/></choice> each correspond directly to a Leiden sigil but make the editor's judgement machine-readable. Gabriel Bodard puts it precisely: "there will always be a one-to-one equivalence between Leiden codes and markup features in the EpiDoc guidelines." Bodard 2010, 6
蒙森第三节默认每一行须记录三事：石上之所有、编者所补、编者之确信程度。印本一次至多承载其二。自 2001 年 EpiDoc-Aphrodisias 项目开始、由 I.Sicily、Telamon、IRT、IRCyr 等所采用之 EpiDoc，即 TEI-XML 之定制版，三者皆得编码。<supplied reason="lost">、<gap reason="lost">、<unclear>、<choice><reg/><orig/></choice> 各直接对应莱顿系统的一个符号，且使编者之判断机器可读。Gabriel Bodard 之表述精确："莱顿编码与 EpiDoc 指南中之标记特征之间，永远存在一对一的等价关系。"Bodard 2010, 6
Bodel, Prag and Roueché argue in their 2024 essay that "the largest current obstacle to digital epigraphy is no longer technical but cultural: our academic culture has not caught up to advancements in technology that make possible the dissemination and sharing of information with unprecedented ease." Their proposed remedy is an "inversion of the existing corpus model", away from a single monumental CIL-like corpus toward a proliferation of local and thematic corpora linked by shared standards, especially EpiDoc and the FAIR principles (Findable, Accessible, Interoperable, Reusable). Bodel/Prag/Roueché 2024, 1, 27
Bodel、Prag 与 Roueché 在 2024 年论文中论道："数字铭文学今日最大之障碍已不再是技术性的，而是文化性的，我们的学术文化尚未跟上让信息以前所未有之便利分发与共享之技术进步。"他们所提补救方案是"对既有总集模式之颠倒"，从单一巨型 CIL 式总集，转向由共享标准（尤其 EpiDoc 与 FAIR 原则，可发现、可访问、可互操作、可重用）所互连之地方性与主题性语料库的繁殖。Bodel/Prag/Roueché 2024, 1, 27
§9 What Mommsen could not have foreseen, the AI inheritance (2022, 2025)蒙森所不及预见者 ， AI 的继承 (2022, 2025)
In 2022, Yannis Assael, Thea Sommerschield, Jonathan Prag and colleagues trained the deep neural network Ithaca on the PHI Searchable Greek Inscriptions corpus to perform three tasks at once: restoration of lost text, geographic attribution, and chronological attribution. Restoration accuracy reached 62% alone; paired with a historian, 72% (up from 25%). Place of origin: 71% accurate. Date: within 30 years of ground truth. Assael et al. 2022 (Nature 603)
2022 年，Yannis Assael、Thea Sommerschield、Jonathan Prag 等人将深度神经网络 Ithaca 训练于 PHI Searchable Greek Inscriptions 语料，使其同时完成三件任务：阙文恢复 (restoration)、地理归属 (geographic attribution)、年代归属 (chronological attribution)。其单独运作之恢复准确率达 62%；与历史学者配对则升至 72%（人类单独基线 25%）。归属地点准确率 71%。年代距真实值 30 年以内。Assael et al. 2022 (Nature 603)
In 2025, the same team, now including Alison Cooley (author of the standard Cambridge Manual of Latin Epigraphy) and Alex Mullen, extended the approach to Latin in Aeneas, trained on a harmonised EDR + EDH + EDCS corpus. Historians find Aeneas-retrieved parallels useful in 90% of cases; their confidence in key tasks improves by 44%; chronological attribution lands within 13 years of ground truth, less than half the Ithaca margin. Assael et al. 2025 (Nature 645)
2025 年，同一研究队伍，现加入 Alison Cooley（《剑桥拉丁铭文学手册》之标准作者）与 Alex Mullen，把方法扩展至拉丁文，发表 Aeneas，训练于 EDR + EDH + EDCS 之协调语料。历史学者发现 Aeneas 检索的平行例证在 90% 案例中有用；他们在关键任务上之自信提升 44%；年代归属距真实值 13 年以内，不到 Ithaca 误差之一半。Assael et al. 2025 (Nature 645)
Charlotte Roueché's Nature commentary made the historical claim explicit: "in the area of humanities, scholars of the classical world have been surprisingly advanced in their adoption of digital tools." Ithaca and Aeneas are the harvest of forty years of editorial decisions baked into EDH, EDR, EDCS, decisions whose lineage runs directly back to Mommsen's erforderlich. The AI works only because every line in its training data was first an editorial judgement made under the four-clause programme set out in the very memorandum above. Roueché 2022; epistemological-genealogy-revised.md
Charlotte Roueché 在《Nature》之配文将历史性论点说清："在人文领域，古典世界的学者在采用数字工具方面意外地领先。" Ithaca 与 Aeneas 是四十年来固化于 EDH、EDR、EDCS 之中的编辑判断之收成，这些判断之谱系直通蒙森之 erforderlich。AI 之所以能用，是因为其训练数据中的每一行，首先都是依据上文那份备忘录所设的四条纲领作出的一项编辑判断。Roueché 2022；epistemological-genealogy-revised.md
Sources consulted (local mount: documents/epidoc/inschriften)
所参考之资料（本地挂载于 documents/epidoc/inschriften）
- Mommsen, Theodor. Ueber Plan und Ausführung eines Corpus Inscriptionum Latinarum. Berlin, 1847. [Source text used for all German quotations above.]
- Bruun, Christer. "The Major Corpora and Epigraphic Publications." In The Oxford Handbook of Roman Epigraphy, ed. Bruun & Edmondson (Oxford: OUP, 2015), 66–100.
- Buonocore, Marco. "Epigraphic Research from Its Inception: The Contribution of Manuscripts." In The Oxford Handbook of Roman Epigraphy, ed. Bruun & Edmondson (Oxford: OUP, 2015), 21–42.
- Wellmon, Chad. "Loyal Workers and Distinguished Scholars: Big Humanities and the Ethics of Knowledge." Modern Intellectual History 16:1 (2019), 85–126.
- Orlandi, Silvia, Maria Letizia Caldelli, & Gian Luca Gregori. "Forgeries and Fakes." In The Oxford Handbook of Roman Epigraphy, ed. Bruun & Edmondson (Oxford: OUP, 2015), 42–65.
- Hopkins, John North & Sarah E. McGill, eds. Forgery Beyond Deceit: Fabrication, Value, and the Desire for Ancient Rome. Oxford: OUP, 2023.
- Van Groningen, B.A. "De signis criticis in edendo adhibendis." Mnemosyne 59 (1932), 362–365.
- Dow, Sterling. "Uniform Style." American Journal of Archaeology 56:2 (1952), 113–118.
- Bodard, Gabriel. "EpiDoc: Epigraphic Documents in XML for Publication and Interchange." In Latin on Stone, ed. F. Feraudi-Gruénais (Lanham: Lexington, 2010), 101–118.
- Bodel, John, Jonathan Prag, & Charlotte Roueché. "Open scholarship: epigraphic corpora in the digital age." In L'épigraphie au XXIe siècle (Bordeaux: Ausonius, 2024), 91–117.
- Kaše, V., Heřmánková, P., & Sobotková, A. "Division of labor, specialization, and diversity in the ancient Roman cities." PLoS ONE 17:6 (2022), e0269869.
- Assael, Y., Sommerschield, T., Shillingford, B., et al. "Restoring and Attributing Ancient Texts Using Deep Neural Networks." Nature 603 (10 March 2022), 280–283.
- Assael, Y., Sommerschield, T., Cooley, A., et al. "Contextualizing Ancient Texts with Generative Neural Networks." Nature 645 (4 September 2025), 141–148.
- Roueché, Charlotte. "Mind the gap, as AI guesses at lost Greek inscriptions." Nature 603 (10 March 2022), 234–235.
- Synthesised from local mining reports:
chapter4-english-revised.md; epistemological-genealogy-revised.md; epigraphic-method-revised.md; symbolic-system-revised.md.