Compositiones Lucenses and Mappæ Clavicula: two traditions or one? New evidence from empirical analysis and assessment of the literature

Abstract

Information on materials and procedures of painters of the past can be gained from the latest examinations of a painting and its materials and from documentary sources, the change of meaning of which is of prior interest to historians of technical art. This paper is an empirical and theoretical examination of the relationships between the two main early medieval collections of craft recipes, the Compositiones Lucenses and the Mappæ Clavicula. The primary aim of this work is to criticise the current prevalent meaning of the concept of Mappæ Clavicula, and to show that its tradition does not include that of Compositiones: these two traditions, despite sharing two sets of manuscripts, result in two appreciably different texts. The first edition of the eighth to ninth centuries recipe book Compositiones Lucenses (Lucca, Biblioteca Capitolare, 490) occurred in 1739, and the twelfth century exemplar of Mappæ Clavicula’s text about one century later (Corning, Museum of Glass, Phillipps 3715, or Corning manuscript). In the interwar period and particularly after WWII, the Lucca manuscript was predominantly considered to be a member of the Mappae Clavicula tradition, which was regarded as second only to Theophilus’s De diversis artibus, as a written source for the study of medieval technology. ‘Compositiones Lucenses’ and ‘Mappæ Clavicula’ are taxonomic concepts for the classification of medieval manuscripts and texts, the meanings of which we redefine in this paper. In contrast to today’s prevailing approach, we move the focus from two single manuscripts (Lucca 490 and Corning) to two different traditions of witnesses, and from single texts to collections of texts bound in the same codex. The critical section of the paper concerns the most important interpretations of the notion of Mappæ Clavicula, while the positive section draws on three works: the seminal paper by Halleux and Meyvaert (1980s), Baroni’s first critical edition of Mappæ, and the inventory of the manuscripts of the Compositiones tradition by Brun (2010s). In the empirical section we contrast the two traditions and consider two sets of items: twelve manuscripts reveal the internal structure of the Compositiones Lucenses tradition, and nine codices, which transmit both traditions, shed light on how these traditions differ. As a result of the present research, we show that a significant segment of the Compositiones Lucenses tradition is composed of an aggregation of small recipe nuclei, and that this tradition developed regardless of that of the Mappæ. The Mappæ Clavicula and Compositiones Lucenses are two distinct textual traditions and not members of a super-corpus.

Background

From a general standpoint, Mappæ Clavicula and Compositiones Lucenses are the sole technological literary records, which come from the Late Antiquity to the Early Middle Ages. “The Mappae Clavicula is second only to Theophilus’s De diversis artibus as a written source for the study of medieval technology” ([1], p. 14). The historical knowledge of painters’ pigments and the interconnection between painting and other crafts have been greatly improved by current chemical analysis. Nevertheless, this forms only one part of the technical history of materials and techniques, because, according to Joyce Plesters “information about the techniques and materials of painters of the past is gained in two ways, firstly from documentary sources and secondly from the examination of the pictures themselves and the materials of which they are made” ([2], p. 101).

Plesters’ work, credited with having laid the foundations of the field of ‘technical art history’, points out that this field requires the study of historical literary sources and physico-chemical investigations of ‘art/craft’ items [3].

Indeed, technical art history requires reliable literary sources; however, this is not the case for the Mappæ and Compositiones Lucenses traditions. In this paper we show how they have been interpreted differently since the first respective publications of their witnesses and how the prevailing interpretation of these records is far from being conceptually well-founded and coherently articulated. We propose a new interpretation of these dated literary records and provide new evidence for their systematic interpretation.

In the period between Ludovico Antonio Muratori’s first publication of the Compositiones Lucenses from the Lucca 490 manuscript in 1739 [4] and Mary Philadelphia Merrifield’s Original treatises, dating from the XIIth to XVIIIth Centuries in 1849¹ [7], some of the most notable Western recipe books of medieval technical ‘art’/craft literature were edited: for example by Gotthold Ephraim Lessing [8], Giuseppe Tambroni [9],² and Sir Thomas Phillipps [10], who respectively published for the first time Theophilus’ and Cennini’s texts, and the anonymous text of Mappæ Clavicula of a manuscript, which is known as the Corning manuscript today.

The recipe book transmitted by the Lucca 490 manuscript has been given arbitrary and contrasting names, including Compositiones ad tingenda musiva [4], Compositiones Variae [11], and Compositiones Lucenses [12]³; it is the oldest witness within the Compositiones Lucenses tradition. This manuscript has been transcribed many times since Muratori‘s first publication [4], and its various aspects have been studied by competent scholars [12,13,14,15,16,17,18]. Giulia Brun created the first inventory of witnesses within the Compositiones Lucenses tradition [19].

The twelfth century text recorded in the Corning manuscript [10] was formed of a series of aggregations around a fourth century A.D. Alexandrian alchemical text. The translation of this alchemical nucleus into Latin, which probably occurred around the fifth century, created the incongruous title Mappæ Clavicula.⁴ After the first edition of the Corning manuscript [10], an English translation of its text was produced in the 1970s by Cyril Stanley Smith and John G. Hawthorne, who compared it with a few witnesses (Sélestat, Bibliothèque Humaniste 17; Lucca, Biblioteca Capitolare 490; and Klosterneuburg, Stiftsbibliothek W.8.293, see [1], pp. 3–9). The connection to alchemy within the Mappæ Clavicula text was first noted by Marcellin Berthelot ([21], pp. 29–30), and, significantly, by Robert Halleux and Paul Meyvaert [20].

Throughout its entire literature, the title ‘Mappæ Clavicula’ was given three contrasting meanings. Since the date of Sir Thomas Phillipps publication in 1847, and for about 90 years, the phrase ‘Mappæ Clavicula’ referred to the contents of the Corning manuscript, which has three incipits, the first of which reads as follows: “INCIPIT LIBELLUS DICTUS MAPPÆ CLAVICULA” (“Here begins the book called Mappæ Clavicula”). See [1], pp. 5–9 for further information on Sir Thomas Phillipps, and the first and only publication of the manuscript.

The second meaning was inaugurated by the research of Rozelle Parker Johnson on the tradition of the Corning manuscript, which was published in 1935–1937. According to Johnson, the title ‘Mappæ Clavicula’ denoted all the texts of the Corning manuscript, including the contents of Lucca 490, and referred to about 80 witnesses [22,23,24]. Johnson’s concept of a unique collection including the Lucca 490 manuscript is shared by Bernhard Bischoff, an illustrious medieval historian and palaeographer. According to Bischoff, the five best-known medieval recipe books (Lucca 490, Sélestat 17, Corning, Madrid 19, and the Klosterneuburg fragment) have a common source that he refers to as Mappæ Clavicula, because they include similar contents ([25], pp. 277–80). This notion was still predominant in the early 2000s: Tosatti [26], pp. 27–36, Kroustallis [27], pp. 69–70, Clarke [28], pp. 23–26, Pomaro [14], p. 171.

In contrast to the authors mentioned above, the present paper sustains the hypothesis that Mappæ and Compositiones are two different traditions, originating from two different texts, whilst often being included within the same set of manuscripts. Meanwhile, a small but decisive number of manuscripts of the Compositiones are completely independent from the Mappæ tradition.

The third meaning, which was preceded by a growing awareness that the contents of the Corning manuscript were heterogeneous, was initiated by Halleux and Meyvaert’s work [20]. Halleux and Meyvaert identified Mappæ Clavicula as only one part of the contents of the Corning manuscript; that is, solely the alchemical nucleus translated from Greek into Latin, which has also been copied in a few other codices (Sélestat, Bibliothèque Humaniste 17 and Glasgow, University Library, Hunterian 110). The critical edition by Baroni et al. [29] took for granted the new meaning ascribed to the title Mappæ Clavicula: the authors increased the number of witnesses of this tradition, without giving its inventory⁵ (for an updated review of the studies on Mappæ Clavicula, see [30]).

Smith and Hawthorne deemed the contents of the Lucca 490 manuscript and that of Corning, which includes the text of the Lucca 490 too, to be two separate text traditions only on the basis of their geographic distributions, the first in the south of the Alps, and the latter in the north of the Alps ([1], p. 4). However, this theory is unsustainable on the basis of a larger inventory of the Compositiones Lucenses tradition.⁶

For other scholars, Mappæ Clavicula consists of two nuclei, called A and B, one of which includes the ancient alchemical core, and the other the Compositiones Lucenses tradition in its most ancient form as demonstrated by the Lucca 490 manuscript ([31], p. 200).

The aims of the present research

From the 1930s until the 2010s, the notion of Mappæ Clavicula, or the analogous Mappæ Clavicula text-family ([28], pp. 25–26), continued to be heterogeneous, ad hoc definitions, without being grounded in any precise codicological and philological bases.

Conceptually, the notions of ‘Compositiones Lucenses’ and ‘Mappæ Clavicula’ relate to the roots of medieval technical literature: they are classifications of medieval texts and manuscripts, which are themselves the results of translations of Late Antiquity technical materials from Greek into Latin. We conceive of these taxonomic notions by redefining their meanings, whilst additionally highlighting the importance and relevance of the present research for the field of the medieval ‘art’/craft recipe books. For the setting up of a better defined conceptual basis, we refer to the literary findings of Halleux and Meyvaert [20] for their identification of an alchemical nucleus in the Mappæ Clavicula tradition, Sandro Baroni et al. for their critical edition of this nucleus [29], and Brun for her research on the witnesses of the Compositiones tradition [19].

We argue that the Compositiones Lucenses and Mappæ Clavicula are two separate collections of texts, and for this purpose we criticise Johnson’s notion of Mappæ Clavicula in both its dated and recent formulations. In order to reach our goals, the works on which we draw must be integrated with the inventories of the manuscripts of both traditions, and founded upon an operative procedure for identifying the text units of the Compositiones Lucenses tradition. The new interpretations of old data will reveal new pieces of evidence, from the examination of a set of manuscripts transmitting the Compositiones Lucenses tradition, and a second set of codices which include texts from the Compositiones and the Mappæ traditions. The material collected in the present research will be used to check the current hypothesis on the existence of a unique corpus containing both traditions. The scrutiny of the present state of the literature will bring further pieces of evidence as to the differences between the two traditions.

A clear example of a change in gestalt: the case of the Klosterneuburg manuscript

An interpretation of Wilhelm Ganzenmüller’s work can give us a good example of a change in gestalt in the notions of Compositiones Lucenses and Mappæ Clavicula. In 1941, the author collated the text of a newly discovered ninth century manuscript fragment (Klosterneuburg, Stiftsbibliothek, W.8.293) with three other witnesses (Lucca 490, Sélestat 17, and Corning). Because 13 Klosterneuburg text units⁷ were discovered to be absent in Lucca 490 and present in the Corning manuscript, he concluded that the new fragment was an exemplar of the Corning tradition [32].

Ganzenmüller’s procedure draws on a limited set of manuscripts and his conclusion is coherent with their contents. An effective procedure requires precise definitions of the witnesses in terms of manuscripts and texts they include, as well as well-defined procedures for assigning a given text unit to one of the two traditions (see below). However, the Corning manuscript is heterogeneous and includes both the texts from Mappæ and Compositiones (see below); for this reason it is not a reliable reference for separating and identifying the two textual traditions. In the present paper, we move the focus from manuscripts to texts and collections of the same text within different manuscripts. We interpret the Lucca 490 as belonging to a different tradition, on the basis of substantial evidence, and the Corning and Sélestat manuscripts as witnesses of both traditions (see the inventories below).

The 13 Klosterneuburg texts do not correlate with the critical edition of Mappæ [29], because the latter concerns the Latin translation of the old alchemical nucleus only. Moreover, unlike Ganzenmüller, we found concordances with the 13 text units absent in the Lucca 490 manuscript with the Sélestat and four other fundamental manuscripts that were unknown as possible witnesses at Ganzenmüller’s time.⁸ Finally, specific lexical terms used by these 13 recipes refer to the same lexicon of other sections of the Lucca 490 manuscript and of the Compositiones Lucenses tradition, as is exemplified by the terms pandius, cianus, lulacin, ficarin, cinnabarin, iarin, etc. These, on the contrary, are completely absent in the Mappæ Clavicula critical edition. Consequentially, one should consider these 13 recipes likely candidates for the Compositiones Lucenses tradition.

In summary, the attribution of this fragment to the Mappæ or the Compositiones traditions involves different meanings of the same terms or labels, different procedures of assignation of a given text unit to one of the two traditions, and different empirical bases.

The inventory of the manuscripts of the Compositiones Lucenses and its text units

Texts within the Mappæ Clavicula and Compositiones Lucenses traditions are at first sight so intertwined that one is initially given to believe that they are simply different arrangements of the same fragments of text. However, Mappae Clavicula can be distinguished from Compositiones Lucences through more accurate and expansive inventories of the witnesses of both traditions and their text units.

In the Lucca 490 manuscript, the Compositiones Lucenses’s text is transcribed on f. 211v (T and U hands) and ff. 217r–231r (mostly N hand) ([14], p. 155). The f. 211v with the texts De fabrica in aqua and De Malta is included in the gathering No. 28⁶, and is thought to filling a blank space ([14], p. 170); the remaining part of the recipe book is included in the gatherings Nos. 30⁸–31¹⁰ ([14], p. 195).

Should the recipes of f. 211v on the foundations of buildings be excluded from the Compositiones’ text? Do they belong to a separate collection?

The split in the textual body of the manuscript needs to be critically interpreted. Muratori [4] began the transcription from f. 217r, line 25 (i.e. from the recipe ‘De tictio omnium Musivorum’), and omitted f. 211v and the first 24 lines of f. 217r (in total 13 text units). This fact did not go unnoticed; the Archbishop Giovanni Domenico Mansi had already remarked in 1751 that Muratori had forgotten to transcribe this folio ([33], pp. 96–97). The succeeding literature, however, appears inconclusive, so that it was not transcribed by Pellizzari [15], nor by Hedfors [17], but by Burnam ([16], p. 14) and Svennung ([12], pp. 26–27). Johnson’s observations that the two textual sections are contiguous in six manuscripts, seems decisive⁹ ([18], pp. 222, 223; [11], p. 25).

The veneration of the Lucca 490, the most ancient exemplar of the medieval recipe books, likely impeded most scholars’ understanding of its idiosyncratic nature. The Lucca 490 textual consecutio, or sequence of text units, is much more disorganised in comparison with other members of the Compositiones tradition, particularly to the Vatican City, Biblioteca Apostolica Vaticana, Reg. lat. 2079 [34]. The latter should be considered the reference witness of the tradition instead of the Lucca 490, because it is well structured, independent from the Mappæ tradition, and includes 188 texts—of which 149 are already present in Lucca 490.

The text of the Compositiones, originally written in ordinary Greek, was likely to have been translated into Latin in the sixth century A.D. ([35], p. 56). The Lucca 490 manuscript “stand[s] as a landmark in the history of sciences and the arts” ([18], p. 224). Currently, the Compositiones Lucenses tradition consists of 26 codices, spanning from the eighth up to the fifteenth centuries ([19], pp. 257–279). Within the codices of Table 1, we distinguish the following notable subsets:

Table 1 The manuscripts of the Compositiones Lucenses tradition

• i. The first 13 codices include texts from the Compositiones and the Mappæ traditions, but only nine of them are fundamental witnesses (Nos. 1–9); three (Nos. 10–12), having a small number of texts, are included in the sub-class of the Compositiones (Mappæ) fragmentary tradition and No. 13 is a descriptus, obtained by a direct copy from a recognised exemplar.

• ii. Two other descripti (Nos. 14–15).

• iii. Six codices transmit only a text that has been renamed Editio Minor (Nos. 16–21); a further seventh witness (No. 5) is fundamental and also includes the Editio Minor. The Editio Minor is composed of 26 text units, which are largely extracted from the Compositiones Lucenses tradition: this small recipe book is exclusively bound immediately after the De Architectura by Vitruvius [36].

• iv. Three codices belong only to the Compositiones tradition (Nos. 22–24).

• v. Two further witnesses contain a few text units of the Compositiones (Nos. 25–26). They represent, together with the above-mentioned manuscripts Nos. 10–12, the fragmentary tradition of the Compositiones Lucenses.¹⁰

Following systematic comparisons of all text units within the entire collection of the Compositiones tradition, we estimate that they number between 200 and 250 texts. Needing a more precise definition, we may state provisionally and operatively that the text units of Compositiones Lucenses tradition consist of three sets (α, β, γ), which have been categorised according to a progressively diminishing degree of probability (from α to γ) that they belong to a critical edition of text units of the Compositiones:

• α. Let us first deal with the text units recorded in the Lucca 490 manuscript. We updated the latest edition edited by Caffaro of the Lucca 490 manuscript, which records 160 texts: four of these are evaluated as ‘texts of uncertain origin’ (on ff. 223r–223v, Nos. 77–80, [37]), because they are likely to have derived from a different, ancient translation of the same textual material that can be read in Mappæ ([38], p. 47; see [29], text units Nos. xxxvii, xliv, xlii, xlvii). We further divided three recipes,¹¹ and the final result consists of 160 texts belonging to the Compositiones tradition.

• β. The α set is implemented by complementary text units of two manuscripts, i.e.: (i) Città del Vaticano, Biblioteca Apostolica Vaticana, Reg. lat. 2079 (ff. 74r–86r, with the exception of texts De coloribus; De emplastro; Ad cruas faciendas on f. 77r as these are not found in any other manuscript), and (ii) Sélestat, Bibliothèque Humaniste, 17, with the exception of the text on f. 37r, the incipit of which reads: Corpus hominis (an excerpt from Homo bene figuratus by Vitruvius).

• γ. Other textual candidates follow a three-step procedure, which is affected by a probability to reject ‘good texts’ (or type I error) and to accept ‘bad texts’ (or type II error).¹²

The inventory of the manuscripts of the Mappæ Clavicula tradition and its text units

The alchemical textual nucleus of Mappæ originally consisted of a preface followed by 193 recipes for Halleux [39], or 182 for Baroni et al. [29]. However, a significant number of recipes of this core is not extant in any witness, with the exception of the titles (rubrications). The entire list (recensio) of the witnesses of the Mappæ tradition is not given explicitly by the authors of the critical edition who quote only 15 manuscripts, and use three papers by Johnson to confirm that the recensio consists within about 25 witnesses ([38], p. 38, n. 30). A more complete inventory is necessary in order to understand how the two traditions are intertwined. After a closer look at Johnson’s papers [22,23,24], we detected information regarding a further ten potential manuscripts of the Mappæ tradition, making a total of 25 witnesses.

The first 13 manuscripts of the Compositiones’ inventory of Table 1 also belong to the Mappæ tradition, and therefore they are not copied into Table 2 (see below). Two other manuscripts are quoted by the authors of the critical edition, but not used for setting up the stemma codicum (Nos. 14–15). Finally, ten further witnesses (Nos. 16–25) were already recorded by Johnson: some of these, marked by an asterisk, might also contain excerpts from the Compositiones tradition.

Table 2 The manuscripts of the Mappæ Clavicula tradition

To summarise, the entire corpus of the manuscripts of Mappæ Clavicula may be divided into three subsets: the fundamental tradition (the first nine manuscripts of Table 1), one descriptus (No. 13 of Table 1), and the remaining 15 manuscripts named the fragmentary tradition of Mappæ.

From touchstone to an alchemical text via a composite codex: a critical review of Rozelle Parker Johnson’s work

Between the two World Wars, research into medieval ‘art’/craft recipe books occurred not only through the discovery and publication of medieval manuscripts, but also through in-depth analysis that took place along two main lines. On one hand, Daniel Varney Thompson Jr. addressed the study of medieval craftsmanship in his seminal 1935 paper [40], by examining 158 manuscripts between the tenth and fifteenth centuries held in 18 libraries. The product of his work is a source-book of medieval ‘arts’/crafts, which includes 202 subjects and sub-subjects ordered alphabetically from absinthium to yellow lead. On the other hand, and most relevant for the present paper, in the 1930s Johnson explored the witnesses of the Corning manuscript scattered over 25 European libraries. He used all the text units of its 1847 first transcription as a   reference [10], and compared them with the alleged witnesses. In so doing, he discovered 81 witnesses of the Corning, some of which contain large sections, and others just short excerpts [22,23,24].

Johnson’s method has at least three drawbacks: (i) it is mainly an exploratory research, which needs to be reinterpreted; (ii) the use of the Corning Phillips as a reference generates discrepancies due to the highly heterogeneous nature of the manuscript, produced by the overlap of chronologically different texts, and (iii) the one and only transcription of the Corning manuscript is considered obsolete.

Firstly, Johnson’s papers are the result of an ‘exploratory work’, because the author “did not attempt to find more than one recipe on each folio” ([23], p. 77; [24], p. 84), and annotated the recipe and folio in concern, which he photographed with his micro-camera. Unfortunately, Johnson was neither able to collate his materials with the Corning manuscript nor to photograph it because Sir Thomas’ heir refused permission ([22], p. 73).

Two manuscripts quoted by Johnson are not useful¹³ and the remaining 79 do not relate to a single text, but rather to the following combinations of three main texts (the number of references is reported in round brackets):

• De Coloribus et Mixtionibus¹⁴ (42).

• Mappæ Clavicula tradition (4), where the term Mappæ corresponds to the notion introduced by Halleux and Meyvaert [17].

• Compositiones Lucenses tradition, where the title ‘Compositiones Lucenses’ is defined as above (7).

• De Coloribus et Mixtionibus + Mappæ Clavicula tradition (1).

• Mappæ Clavicula tradition + Compositiones Lucenses tradition (6).

• De Coloribus et Mixtionibus + Compositiones Lucenses tradition (9).

• De Coloribus et Mixtionibus + Compositiones Lucenses tradition + Mappæ Clavicula tradition (9).

• Lucca manuscript 490 (1).

Secondly, the Corning manuscript is very heterogeneous, being a compilation of compilations and the result of many different chronological contributions ([1], pp. 7, 15, 17; [39]). It includes the following items quoted in accordance with the enumeration (between brackets) of Smith and Hawthorne’s edition [1].

• i. The recipe book named De Coloribus et Mixtionibus (Nos. i–xi), ff. 1r–4r.

• ii. What is today known as the alchemical core of the Corning manuscript or Mappæ Clavicula (0–95), ff. 4r–24r.

• iii. Texts belonging to the Compositiones Lucenses tradition (Nos. 96–278), ff. 24r–62r. Notable text units of English and Arabic origin are Nos. 190–191 on f. 40r and Nos. 195–203 on ff. 43r–44v. Texts addressing military subjects and incendiary bombs are Nos. 264–278 on ff. 57v–62r.¹⁵

• iv. Table of runes (Nos. 288 B-P on f. 64r).

• v. Some enigmatic notes interpreted as the legend of automata, the pictures of which are missing on f. 65r (see [35], p. 58).

Thirdly, the transcription of the Corning manuscript used by Johnson [10] is not especially reliable because Smith and Hawthorne [1] only split around 92 text units out of 293 of the text units.

Johnson’s last work [11] is a revision of his expanded PhD dissertation on the Lucca 490 manuscript. His tragic death due to a car accident in 1941 interrupted his seminal scientific research (see Johnson’s obituary 1899–1941, [44]). Johnson, and a great majority of authors after WWII, shared the conviction that there was a high level of continuity in the literary technical materials between the Late Antiquity up to the fourteenth to fifteenth centuries ([11], pp. 88, 89; [21], pp. 16–17; [22], pp. 72–73), and for that reason Johnson considered the Lucca 490 manuscript as a witness of the textual tradition of the Corning’s text ([24], p. 84). Thompson had similar ideas when he published the transcription of the fourteenth century manuscript De Coloribus Illuminatorum Siue Pictorum. Thompson was convinced that this manuscript stood in a long line of recipe books beginning with the Lucca 490, and continuing through to the Schedula of Theophilus and the Corning’s text ([45], pp. 280–1). Much later, Inès Villela-Petit formulated an opposing opinion on De Coloribus Illuminatorum ([46], pp. 172,174).

Mark Clarke’s approach to the early medieval recipe books is likely to have been determined by the same idea of continuity, and does not consider the hypothesis that a further independent tradition from that of Mappæ could exist. His definition of the Mappæ Clavicula text-family ([28], p. 25), as the union of four manuscripts (Lucca 490, Madrid 19, Corning, and Sélestat 17), is in terms of manuscripts and not of texts as in the present work. In fact, the four items are witnesses of both traditions, and all four are of fundamental importance to the reconstruction of Mappæ and Compositiones texts. Clarke’s list lacks five other fundamental manuscripts (see below): he follows the paths of Bischoff and Johnson, and his method is affected by the similar and even greater flaws. In fact the four reference manuscripts contain different texts, a great part of which is irrelevant to either of the respective traditions.

The textual arrangement of the Compositiones Lucenses tradition

The contents of 12 manuscripts of the Compositiones Lucenses tradition show the presence of at least ten thematic nuclei of text units.¹⁶ Table 3 illustrates the presence of six nuclei, which are included in four notable manuscripts dated between the ninth and the twelfth centuries. Usually, every manuscript includes a part of the nuclei presenting a peculiar overall arrangement.¹⁷

Table 3 The arrangement of six thematic nuclei in four witnesses of the Compositiones Lucenses tradition

In order to obtain Table 3, we firstly indexed every text unit of each manuscript with a progressive number, so that it became possible to obtain the consecutio (the sequence or ordinal structure)¹⁸ of the text units within each of the four manuscripts examined here and their respective corresponding folios. The first column of Table 3 refers to the name of the textual nucleus, and the second to the title of the text units, while columns 3–6 indicate the relevant folio and the position of each text-unity within its own manuscript.

The nuclei are internally ordered in an even fashion. Table 3 provides examples of the change mechanisms which occurred in the four manuscripts: the most common of which is the loss of one or more than one text within a nucleus. Other mechanisms include the lack of an entire nucleus (see the case of lapidary ‘b’) and the loss of the ordinal structure, due to inversions. In a few cases, these nuclei are discontinuous, which is probably due to the conditions of their exemplar or to a confusion between the folios during the copying process.

Some codicological information on the main manuscripts of the Mappæ Clavicula tradition

In the current section, we outline the main pieces of codicological information on the nine fundamental codices, which were used for the critical edition of Mappæ’s text (ninth to fifteenth centuries; see [47]). The stemma codicum of Mappæ (Fig. 1) is composed of two branches, both descending from the same lost manuscript. To sum up, the α family is characterised by the transmission of title, prologue, and index, whereas the β family lacks the prologue, index, and the first four texts of Mappæ Clavicula.

Fig. 1

(Source: Courtesy of Sandro Baroni and Paola Travaglio ([36], p. 39))

Stemma codicum of the Mappæ Clavicula tradition

Two different selective forces operated on the two branches of the stemma. The first is represented by the selective action of the various copyists, who introduced a set of sharp changes to the Mappæ’s texts of the α family along the chain from ω to α₄. The active selection operated on the β family was that of the copyist who compiled the florilegium.

The structure of the texts within the codices may be described with the mechanism pair ordinatio/compilatio.¹⁹ A palaeographical examination of the nine codices reveals that the various scribes did not use any particular visual devices to organise the material in any specific way. Ordinatio’s indexes, numeral or in alphabetical order, marginal numbers or glosses are all features of an efficient textual mise-en-page. Nevertheless, these elements are missing in our codices, with the exception of the London, British Library, Add. 41486 manuscript (see [38], p. 43). The texts of the eight remaining codices are provided in a continuous progression, in which recipes are marked by their title only, sometimes with differently coloured ink, and without any comments in the pages’ margins. A discontinuous alpha-numerical thematic order is rarely present.

The statistical description of the textual blocks of the nine fundamental manuscripts

Figure 2 illustrates how the codices have been compiled with sets of text units of various magnitudes, indicated by blocks of different colours. The relevant folios and quantity of texts are specified inside each block. In addition, the greenish blocks include an integer within round brackets, which denotes the quantity of the text units already present in the Lucca 490 manuscript. The block of texts of the Mappæ and Compositiones traditions, which are distinguished in the α and β families, are considered the material source for a statistical treatment: the unit of analysis are the blocks of texts belonging to the two traditions.

Fig. 2

Contents and structure of the nine codices transmitting Mappæ Clavicula and the Compositiones lucenses traditions. α Family—S Sélestat, Bibliothèque Humaniste, 17, tenth century. Od Oxford, Bodleian Library, Digby 162, thirteenth century. L London, British Library, Add. 41486, thirteenth century. C Corning, Museum of Glass, Phillipps 3715, twelfth century. Ob Oxford, Bodleian Library, Bodley 679, thirteenth century. Om Oxford, Magdalen College, 173, fourteenth century; β Family—M Madrid, Biblioteca Nacional, 19, twelfth century. P Paris, Bibliothèque National de France, lat. 7418, fourteenth century. Fp Firenze, Biblioteca Nazionale Centrale, Pal. 951, fifteenth century. Green Compositiones Lucenses tradition. Red Mappæ Clavicula. Red and grey fragmentary tradition of the Mappæ Clavicula. Grey other unrelated texts

The four columns on Table 4 present four univariate, discrete sets of data, small in size: they refer to the number of text units indicated by the blocks of Mappæ (Compositiones) of the α and β families.

Table 4 The statistical description of the blocks of texts of the α and β families

It seems hard to consider the four data sets as samples extracted from one large population of codices; therefore, we may consider each one of the four as individual populations themselves. The average number of texts of the Compositiones of the α family (98.5.) is 1.72 times that of Mappæ (57.2), unlike the figures of the β family which are very close in number (26.7 texts for the Compositiones and 24.6 for Mappæ respectively).

The coefficient of variation (CV, the ratio of the standard deviation to the mean value) of the four distributions is large and has similar magnitudes. It ranges from a maximum CV of 0.85 (Mappæ blocks of the α family) to a minimum CV of 0.67 (Compositiones blocks of the β family). The high values of the standard deviation and the coefficient of variation of the four distributions are due to the contemporaneous presence of both big and small blocks of recipes.

The indices of kurtosis and skewness are used to describe the shape of the distribution of our empirical populations. Karl Pearson operatively introduced the concept of kurtosis in terms of the fourth moment around the mean, and coined the terms leptokurtic, mesokurtic, and platykurtic to indicate cases in which (excess) kurtosis is > 0, = 0 (K value of column 2 is near to zero), and < 0 (columns 1, 3 and 4), respectively. This index is related to the tails of the distribution: high values of kurtosis mean that the distribution is affected by infrequent extreme deviations, that determine a great part of the variance, This is the case of the Compositiones distributions of both α and β families (see columns 1 and 3).

The skewness index has been calculated in terms of the third moment around the mean (Fisher–Pearson standardised third moment): in symmetrical distributions, the mean is equal to the median, and the distribution has zero skewness. Positive skewness indicates that the tail on the right side of the distribution is longer or fatter than that of the left side, and the mass of the distribution is concentrated on the left of the figure. Distributions nos. 3 and 4 are moderately skewed to the right (+ ½ < skewness < + 1), while nos. 1 and 2 are approximately symmetrical (0 < skewness < + ½). According to the Shapiro–Wilk index, three distributions over four are normal at the confidence level of 0.05, and only data set no. 3 is non-normal.

The mean values of the two sets of data including all the blocks of Compositiones (columns 1 plus 3) and of Mappæ (columns 2 plus 4) are 65.00 texts (Compositiones) and 41.66 texts (Mappæ). Their corresponding coefficients of variation increase in comparison to the values of the four initial data sets [CV (CLT) = 1.07; CV (MCT) = 0.99]. The high values of the CV may be easily interpreted as the result of an active selection of text units by the copyists. The Shapiro–Wilk index shows that both new distributions are normal at 0.05 level of confidence.

The codices of the α family

In the α family, Mappæ Clavicula’s text always precedes that of the Compositiones Lucenses. No specific devices—such as title, spaces, transcription from a new folio or fascicule, etc.—separate the texts of the two traditions, with the exception of the Sélestat 17. All manuscripts which succeeded the lost manuscript α₂ lacked the last section of Mappæ (recipes Nos. CXLVII–CLXXXII, [38], p. 41; see Fig. 1).

[S] Sélestat, Bibliothèque Humaniste, 17, tenth century, Saint Amand of Tours in Northern France, parchment, 215 ff., 20–22 lines in a single column. The codex is a collection of texts on craft techniques and architecture, and is likely to be the result of a subject-based reassembly of different writings. Texts from Mappæ and Compositiones traditions are bound with De architectura by Vitruvius, De diversis fabricis architectonicae by Faventinus, excerpts from De re rustica by Palladius, and other fragments from ancient collections on the same subject.

[Od] Oxford, Bodleian Library, Digby 162, thirteenth century, England, parchment, 48 folios, 46 lines in two columns. It is the most extensive Mappæ witness within the α family. Likely, Od is a late copy of an ancient version of the text, as it represents the only witness of the α family, which includes five texts on ff. 21r–21v (Nos. CXXVII, CXLII, CXLIII, CXLC, CXLVI of the critical edition, see [38], p. 42).

[L] London, British Library, Add. 41486, thirteenth century, Italy, parchment, 252 folios, 25 lines in a single column, 14 folios and two entire quires missing. This manuscript consists mainly of alchemical texts and recipe books. A single, competent compiler, who composed glosses and word lists, seems to have efficiently gathered and organised texts (De diversis artibus by Theophilus, De coloribus et artibus Romanorum by Heraclius, De Coloribus et Mixtionibus, Mappæ Clavicula and Compositiones Lucenses. See [38], p. 43). The prologue of Mappæ is placed at the end of the codex (ff. 221v–222r), together with those of Heraclius and an incomplete version of De Coloribus et Mixtionibus.

Three texts on ff. 11r–11v, indicated here as ‘text of uncertain origin’, could be ascribed to Mappæ Clavicula; however, they seem likely to be part of autonomous traditions, excerpted from Mappæ or from other similar alchemical collections. Two Mappæ texts on f. 74r have been duplicated on f. 60v. The manuscript is dismembered after f. 100v. This is indicated by an interruption in the transcription of the text De speciebus metallorum herbarum lapidum lignorum et fusi salnitri afronitri at the end of the quire. Folio 101r begins with a text on nigellum.

[C] Corning, Museum of Glass, Phillipps 3175, twelfth century, England, parchment, 67 folios, 21 lines in a single column: its texts (see above) are not separated by particular visual devices.

[Ob] Oxford, Bodleian Library, Bodley 679, thirteenth century, monastery of Saint Augustine in Canterbury, parchment, 140 folios, 50 lines in a single column. The manuscript is mostly devoted to alchemy and recipe books such as De Coloribus et Mixtionibus, Mappæ Clavicula, and Compositiones Lucenses. The compiler carried out an extensive and recognisable reworking of the text. The section on ff. 23r–32v contains several textual segments from the Compositiones tradition: some of them could have been missing, due to the loss of two folios (ff. 33–34).

[Om] Oxford, Magdalen College, 173, fourteenth century, England, parchment, 264 folios, 27–33 lines in a single column, ff. 79r–95v in two columns. It is an alchemical and medical codex that includes texts from the De Coloribus et Mixtionibus, Mappæ Clavicula, Compositiones Lucenses tradition, with excerpts from the first book of De diversis artibus by Theophilus. One text unit from Mappæ is inserted within the transcription of the Compositiones tradition.

The codices of the β family

Madrid 19 still maintains a division between the blocks of texts of the two traditions (see Fig. 2); the other two codices are characterised by an alternation between large and small segments of text units of the Compositiones and Mappæ traditions. It is plausible that all manuscripts of this family come from the same florilegium, even though this has only been demonstrated for Madrid 19 and Paris, lat. 7418 [49]. The β family is characterised by the presence of two typical recipes.²⁰

[M] Madrid, Biblioteca Nacional, 19, twelfth century, Catalonia or Montecassino, parchment, 203 folios, 47 lines in two columns. This codex originates from a florilegium, compiled in Southern Italy around the year 1000. Some texts on ff. 199r–202v are named in the library’s catalogue ‘Excerpta ex tractatibus alchimiae’: they belong to Mappæ and Compositiones traditions. The manuscript includes the most extensive copy of the alchemical nucleus of Mappæ within the β family ([38], p. 45), which is preceded by the recipe Item de chrisographia on f. 199ra; the other typical recipe Aurum crescere is on f. 201rb.

[P] Paris, Bibliothèque National de France, lat. 7418, thirteenth to fourteenth centuries, Italy, parchment, 284 folios, 42 lines in a single column. This manuscript belongs to the same florilegium of Madrid 19, with which it shares most features, although the Parisian manuscript appears more extensive than the Madrid codex in the section of the Compositiones tradition. The texts Item de grisografia and Amplificatio auriis are on f. 269ra, and f. 278rb respectively.

[Fp] Firenze, Biblioteca Nazionale Centrale, Palatino 951, fourteenth to fifteenth centuries, Southern Italy, paper, 25–26 lines in a single column, ‹dragon› and ‹M› watermarks alternated in the volume. This manuscript is most likely to be an excerpt from the same florilegium, although here Mappæ Clavicula and Compositiones traditions are bound up with De diversis artibus by Theophilus, and other treatises related to pigments and glass colourings. It shares with the P manuscript the presence of the recipe De grisographia on f. 17r.

Some codicological information on three witnesses of the Compositiones Lucenses tradition

The following three manuscripts belong only to the Compositiones Lucenses tradition: they are fundamental witnesses of the tradition, two of these include large amounts of text units, and two are very old. All three are independent from the Mappæ tradition.

[Lu] Lucca, Biblioteca Capitolare, 490, eighth to ninth centuries, Lucca (Italy), parchment, 32–34 lines in a single column, except 18 lines on f. 211 (see above).

[V] Città del Vaticano, Biblioteca Apostolica Vaticana, Reg. lat. 2079, twelfth century, France (Rouen?), parchment, 39 lines in a single column. The text of the Compositiones Lucenses tradition was copied just after the transcription of De architectura by Vitruvius. It begins with a series of directions on incendiary mixtures and military machines, and this fits in well with the Book X of the De architectura devoted to the use and construction of machines.

[K] Klosterneuburg, Stiftsbibliothek, W.8.293, ninth century, France, Belgium or Germany, parchment, 27 lines in a single column, 30 texts. This fragmentary manuscript was probably formed from at least 119 text units.

Results

The catalogue of the library of the Reichenau’s monastery (821–2), and the title transmitted by the Sélestat 17 manuscript (tenth century), indicate that the phrase ‘Mappæ Clavicula’ has been around since the Carolingian times ([20], p. 11). Various pieces of evidence show that Mappæ and Compositiones are two different textual traditions, arising from new interpretations of old evidence, an assessment of the literature, and from new codicological data.

The text of Mappæ can be retraced throughout the indexes of four manuscripts.²¹ The Compositiones Lucenses’ text, instead, cannot be restored on the basis of current knowledge, since none of its manuscripts identified to date can give us the precise number of text units and their sequence.

The topics dealt with by the two traditions diverge significantly. Mappæ Clavicula has a distinctly alchemical character and only briefly touches upon craft applications, as highlighted by both its enigmatic prologue and the sequence of its texts. In fact Mappæ Clavicula was originally a coherent work, its subjects arranged in reasoned order: according to its critical edition ([38], p. 41), the text units of Mappæ are organised, with some exception, in order of metal from precious to cheap. The progressive number of the text units (in brackets) are the following: gold (Nos. 1–70); silver (Nos. 71–113), copper (Nos. 114–136); iron (Nos. 142–146); lead (Nos. 147–151), tin (Nos. 153–155); and glass (Nos. 158–172). Glass was considered a metal by Late Antiquity astrologers and alchemists, because it melts like other metals (see [50], pp. 136, 149–160, 162).

In stark contrast, the contents of the Compositiones tradition are specifically devoted to craft techniques, without any substantial mention of alchemical or hermetic themes, as has been noted by Berthelot ([21], pp. 10, 22).

Besides, there are some lexical divergences about pigments and dyes between the Corning and the Lucca 490 manuscripts, which are likely to be mirrored in other witnesses to the two traditions. What the first manuscript calls aerugo (copper(II) acetate), indicum (indigo), caeruleum (Egyptian blue), alcusa (alkanet), aes ustum (copper (I) oxide), chalcanthum (iron and copper sulphates), the second manuscript calls iarin, lulacium, quianium, lacca, cecucecaumenon, and bitriolum respectively ([39], pp. 179–180).

Pieces of codicological evidence

Two pieces of evidence show the separation between the two traditions. Firstly, a significant part of the Compositiones Lucenses tradition was likely to result in the aggregation of small internally ordered nuclei; each manuscript presents its own particular order of these nuclei that form the Compositiones Lucenses text (see also [39], p. 178). This feature sharply distinguishes this tradition from that of Mappæ (see [27] for an opposing opinion).

Secondly, three witnesses of the Compositiones tradition had their autonomous literary circulation from the Mappæ tradition, from at least the ninth century until the twelfth century (the eighth to ninth centuries Lucca 490, the ninth century Klosterneuburg W.8.293, and the tewlth century Vatican, Reg. lat. 2079).

Three other notable pieces of evidence concern the Mappæ Clavicula tradition; i.e.:

• i. The Corning manuscript is a poor candidate to reconstruct the contents of Ur-Mappæ. Its position in the stemma codicum of Fig. 1 is low. It is extremely heterogeneous, and its first and unique transcription [10] is not reliable (see above)].

• ii. Unlike the Compositiones, the Mappæ tradition is not characterised by witnesses having both greatly sized blocks of texts and an independent life from other textual traditions. Based on what we currently know, only a small number of manuscripts of the Mappæ’s fragmentary tradition might be independent from the Compositiones Lucenses.

• iii. Any theory on Mappæ should consider that the texts of the fundamental tradition of Mappæ have been subjected to sharp changes. The first and greatest change is the separation of the α and β families of the Mappæ’s text, which have different philological ([38], pp. 38–40) and statistical features. Moreover, they are transmitted with different mechanisms: the three codices of the β family are more conservative, being members of a florilegium ([38], pp. 45–46), whereas the α family is strongly affected by a set of abrupt consecutive changes from α₁ to α₄. Twenty-nine texts are missing from the text change α₁ to α₂ ([38], p. 40). Similarly 36 texts are missing in the transition from text α₂ to α₃ ([38], p. 41). In the transition from α₃ to α₄, texts numbered with Nos. 121 onwards are missing ([38], pp. 42–43).

The analysis of the manuscripts which include both traditions reveals two further pieces of evidence. Firstly, the codices’ structures (see Fig. 2) indicate a variety of different arrangements. We focus primarily on large blocks of text units: the biggest number of texts of the blocks of the Compositiones (Mappæ) tradition varies from 249 (127) texts in the α family to 55 (65) in the β family. Four codices out of six of the α family (S, L, C, and Ob, see Fig. 2) include big blocks of text units from the Compositiones’ tradition (116, 176, 249, 123 texts), and in three cases out of four (L, C, and Ob) the texts of the Compositiones’ blocks overlap those of Mappæ Clavicula, without the two traditions becoming mixed in any way. It seems safe to conclude that the two traditions overlap in many cases of the α tradition. Moreover, it must be noted that the two Compositiones blocks of the Sélestat codex are practically complementary (see Fig. 2). These features may be rationalised with the deliberate copying of a selection from larger blocks operated by copyists. Nevertheless, the origin of big blocks of texts may also be explained by means of a first stage of aggregation, followed by a successive segregation from a unique source in a quasi-homogeneous state from a source named Mappæ Clavicula text family. However, the latter hypothesis is not statistically plausible.

Secondly, with the exception of the Lucca 490 manuscript and sparse information on some of the nine fundamental codices, our knowledge of their codicological units (quires or booklets) is still approximate. Nevertheless, further evidence of separation of the two traditions emerges: in the Sélestat 17 manuscript, the Mappæ Clavicula text is found in the first two quires, the second of which (ff. 10–13) is incomplete, as four folios are missing, and the last page is largely blank (f. 13v). Compositiones Lucenses tradition starts from a new quire on f. 14r, which is different in both format and parchment. The two traditions are contiguous and are copied in subsequent gatherings. This same fact may be interpreted as a process in which two different texts or booklets are bound together, and not as the result of the deliberate work of the copyist. However, this latter interpretation cannot rule out the magnitude of the two Compositiones’ blocks, neither their complementarity.

Conclusions

It is doubtful that a critical edition of the Compositiones Lucenses tradition will be possible. However, a comparative and perhaps critical edition of the text-units of the latter seems plausible. This will contribute to a more clear-cut distinction between the two traditions.

In summary, there are new pieces of evidence supporting the theory of two separate collections of recipe books. First of all, any theory on Mappæ must consider its critical edition, which heavily narrows the number of possible witnesses and the fact that the Compositiones Lucenses tradition has had an autonomous existence from Mappæ for a long time. The contents of the two traditions are significantly diverse and Compositiones’ structure includes various nuclei of texts, the contents of which are not similar to those of Mappæ.

Secondly, other pieces of evidence emerge from an analysis of the structure of the witnesses with texts of the two traditions: we refer to the large sized blocks of texts of Compositiones in the α family of Mappæ, the statistical differences between the two traditions as well as the differences between the α and β families. Moreover, the copying process of the nine fundamental codices occurred in various manuscripts through a simple superposition of segments of different, but significant sizes. In the Sélestat 17 and other manuscripts of Fig. 2 there is a clear-cut separation between the two traditions. Further evidence could come from assessment of the lexical differences on pigments and dyes throughout a significant number of witnesses from the two presumed traditions.

All these facts are difficult to account for with any theory of a unique Mappæ Clavicula text family; rather, they can be rather easily explained by imagining the existence of two separate traditions.