This article attempts to shed new light on the collaboration between the chymical authors Michael Maier (1568-1622) and Francis Anthony (1550-1632) using stylometric authorship attribution. Maier and Anthony were friends and we know that they worked together on the English and Latin versions of Anthony's Apologie or Apologia (1616) respectively. The question remains whether Maier was more than just a mere translator, as it has been claimed in the past - notably by Maier himself. Using R-Stylo, stylometric analyses are conducted. It is discussed what conclusions can be drawn from them given that we already know Maier and Anthony were working together and that Maier was the translator responsible for the Latin Apologia (1616) ascribed to Anthony. In the end, stylometry doesn't ofer enough evidence for us to make any definite claims regarding the authorship situation under discussion. It can, however, ofer certain insights into the stylometric proximity between Maier and Anthony.
and suggestions for the use of Anthony’s drinkable gold. The arguments made in the respective
ifrst parts seem identical upon cursory comparison. Thus, the diferences investigated by Lenke
et al. to establish who contributed what pertain mainly to the frontmatter. This assertion,
however, is in stark contrast to the opinion of Lenke et al. that “[the Latin Apologia and
the English Apologie] are, in fact, quite separate works with substantial variations in content”
[18, p. 7]. The frontmatter, especially the dedicatory pages, were often subject to last-minute
changes in times of hand-press print, as can been seen in alchemical book-making practices:
Piorko shows on the example of a ‘ghost issue’ of Arthur Dee’s Facisculus Chemicus (1631)
that the exact wording of hand-press books wasn’t necessarily fixed across diferent issues of
the same edition [
Consequently, this paper aims to ofer a diferent approach to the problem outlined above. Rather than comparing the contents of the English and Latin versions of the apologetic tract, it sets out to compare both Latin works of Anthony’s stylometrically – amongst each other and to a set of other works authored by Maier. Lenke et al. have suggested that, unlike the Assertio with its simpler language, the Latin Apologia corresponds to Maier’s elegant prose [18, p. 6].4 In contrast to the hard-to-determine question of compositional priority, the question of style can be tackled by digital stylometry. Accordingly, this short paper presents and discusses the results of authorship attribution algorithms run on the texts in question in a closed-game scenario.
The stylo R library was used. Given the highly inflected nature of the Latin language, Eder’s
Delta was chosen [
3This is likely because the main matter doesn’t engage with the Rosicrucians at all – it is an allegorical Tierstreitgedicht. The purpose of the preface is to frame the text as Rosicrucian even if the text itself might originally not have been intended as such.
4As an addendum it is important to note that ‘elegant prose’ is a quite vague notion used by Lenke et al. which is hard to quantify unless we establish a set of criteria for what counts as ‘elegant’ or not. It is questionable whether stylometry can trace or reproduce such categories in an adequate manner or whether they even make sense at all. 500 and 3000 most frequent words. The results of both covariance Principal Component Analysis scatterplots as well as dendrograms resulting from the Eder’s Delta analysis were taken into account for interpretation.
The corpus of works to be compared contained the Assertio (1610) which is clearly attributed
to Francis Anthony, the Latin Apologia (1616) whose authorship remains to be investigated as
well as a number of publications clearly authored by Michael Maier.5 None of those texts is
shorter than 5000 words which is deemed to be a critical point for successful attribution [
8The expression ‘entirely garbled sections’ refers to OCR artifacts produced in the NOSCEMUS model when ornaments in print are mistaken for text, often resulting in Ancient Greek junk text in the outputted removed, too, as these are highly likely to be junk.9 However, the analyses were still run on mostly dirty OCR which, despite the high performance of the NOSCEMUS General Model, contains a number of imperfections and probably still afected the accuracy of the analyses in transcription. This is to say that no actual text was lost in the analysis, except possibly for the occasional missed line in the Layout Analysis stage of the Transkribus process. This does not amount to any substantial loss of text material.
9These mainly come from (snippets of) catchwords, i.e. the first word or part of a word from the following page, set at the bottom of early modern printed books to facilitate book binding. These often come out incorrectly or are half-missing in transcriptions resulting from HTR+ with the Noscemus model. Since they aren’t relevant to our analysis, they can safely be removed. one way or another. To counterbalance the efect of the dirty OCR which contains some nonexistant words due to incorrect transcription, the decision was made to use a higher number of most frequent words than the 50–150 which are often cited to provide good results.10
Once the corpus had been prepared, Principal Component Analysis and Eder’s Delta using
the 500 most frequent words were first conducted on a corpus also containing two works not
authored by neither Anthony nor Maier, Augurelli’s Chrysopoeia (1515) [4] and Johnson’s
Lexicon Chymicum (1657) [16]. This step was included to check whether Anthony’s publications
are stylometrically close to Maier’s at all. Compared to the two other chymical authors,
Johnson and Augurelli, both of Francis Anthony’s alleged works are quite close to Michael Maier’s
corpus (see figures 1 and 2). Since Augurelli and Johnson are classified as so very diferent
from Maier and Anthony, the distinctions between the latter get blurred in the analysis.
Consequently, those works were removed from the corpus in a second step to make diferences in
the ‘inner circle’ of Maier and Anthony’s works more visible (see figures 3 and 4). Since it
wasn’t clear whether the genre diferences – Augurelli’s work is poetic, Johnson’s is a lexicon
– would negatively impact the analysis, a third foreign author was later introduced for
evaluation purposes: Arthur Dee’s Facisculus Chemicus (1631) [
10While a typical amount of most frequent words taken into acount in the classical Burrow’s Delta measure is about 50–150, the efectiveness of using a higher amount of words has been investigated in multiple studies. It remains unclear, however, if using more high frequency vocabulary is a good way of compensating for OCR imprecision or how much the slight OCR imprecision even afects the stylometric results – probably not at all according to Eder [9, p. 612]. Alchemical content words appear in our corpus starting at around the 400 most frequent words, so the decision was made to go with the 500 most frequent words to include some of those alchemical keywords which might indicate thematic tendencies. The texts under scrutiny are all fairly similar in terms of content after all.
11For the ‘closed game’ analysis without Dee, Augurelli and Johnson we are at 546083 tokens, with just Dee added at 563812 tokens and at 634390 tokens for the analysis with all authors included.
Tweaking the parameters of the analyses (like the number of most frequent words) causes the clusters of the PCAs and the branches of the dendrograms to shift around a little but the overall groupings remain similar irrespective of the number of most frequent words. Especially the mythalchemical and encyclopedic works in the style that Maier is famous for – Arcana, Symbola and Septimana – remain closely connected throughout all analyses. One can also observe that Maier’s works wherein he is more dependent on other authors, such as his collection of three texts translated to Latin by Maier, the Tripus Aureus [27], and his Examen Fucorum [23] in which he reuses arguments by Heinrich Khunrath (1560–1605) from his Trewhertzige WarnungsVermahnung eines getrewen Liebhabers der Wahrheit (1597) tend to cluster together [5]. The same goes for works in which Maier discusses more practical matters, mostly concerning metals and iatrochymical medicines, such as Coelidonia, De Circulo Physico, Viatorium and Civitas Corporis. Those unsurprisingly also show up in closer proximity to Anthony’s Assertio, a Paracelsian tract praising his golden medicine (aurum potabile).
Earlier it was stated that 40–80% culling has shown more promising results. This can be seen in the respective figures 6 and 7 which show the results of analyses run with Eder’s Delta, 500 most frequent words and either with just Dee’s text added or with all texts taken into consideration including the ‘control group’ of Augurelli and Johnson. In the analyses without culling, it was surprising that neither Maier’s Examen – which contains substantial amounts of Latinized paraphrasing of Khunrath – nor his Tripus Aureus – a collection of three translations by Maier – were classified as substantially diferent from Maier’s other works. If the translator really is invisible in stylometry [31, p. 231], why are Maier’s translations (Tripus Aureus or the Apologia) not outliers to the rest of the corpus? Forsyth and Lam state that authorial discriminability should be preserved in translation and posit the “working hypothesis that both the author’s and the translator’s ‘handprints’ are present in a translated work and, with suitable tools, both can be revealed” [13, p. 213]. If we follow this hypothesis, we should see some efect of the original texts and their authorial handprints ‘foreign’ to Maier in both the Examen and the Tripus. In the analysis without culling, this was not noticeably the case. When re-run with the above-mentioned culling, however, the analysis shows these two texts to be more clearly separate from Maier’s other works as well as better distinguishing the works foreign to Maier’s corpus. This seems more reasonable from a historiographical perspective, despite the fact that an earlier text of Maier’s, the 1609 Coelidonia, now gets mixed into the branch of other authors (see figures 6 and 7). Given that Maier’s Coelidonia deals with iatrochymical medicine and chrysopoeia and thus being very close to Anthony in content, we may assume that content words – or maybe diferences in style related to talking about certain topics – might have some sort of influence on the authorship attribution after all.
In figure 7 (PCA), we witness a diference in author signal between Anthony’s Assertio – which is grouped in with some of Maier’s texts focusing more on metals – and Anthony’s Apologia which appears conspicuously close to Maier’s texts authored around the same time. With the culling applied, Anthony’s texts appear very close together and seem more separate from Maier’s in the PCA, yet – unlike, for instance, Dee’s text – still get grouped into the ‘Maier branch’ in the dendrogram (see figure 6). In the end, it has to be noted that all texts from the ‘alchemical prose’ group are fairly close together, so it’s ultimately hard to tell whether we should draw strong conclusions from those results.
Initially we asked the question: How close is the Latin Apologia to Michael Maier’s published works? Surprisingly close. However, given that Anthony’s Assertio was also fairly close to Maier’s other works stylistically, the question remains how meaningful this result really is for historical interpretation.
Rybicki asserts that “Delta usually fails to identify the translator and identifies the author
of the original instead” [31, p. 231]. It follows that the translator should be stylometrically
invisible. However, in our case, the clusters of Maier and Anthony were quite close to each
other. Could this indicate Maier’s involvement in both books? The historical circumstances
make this very unlikely in the case of the 1610 Assertio. What else then could be the reason
for this stylometric result? Maier and Anthony were friends who bonded over their shared
interests, primarily in drinkable gold as chymical medicine. It thus probably isn’t surprising
that they are grouped closely together in a stylometric analysis which is influenced by content
or in the very least ‘content-conscious’ [
On the one hand, Maier and Anthony seem to be overall quite close to each other in terms of content. On the other, the Latin Apologia (1616) – where Maier’s involvement is suspected – is closer to Maier than the Assertio (1610). If we accept Rybicki’s conclusion of the translator’s stylometric invisibility [31, p. 246], it follows that the closeness of the Apologia to Maier’s opus might indicate some kind of involvement on his part which goes beyond translating, irrespective of the dubious argument concerning the diferring frontmatters outlined above. Further analyses need to be done if more certainty on this matter is to be achieved. However, it remains questionable if a clear demarcation of who contributed what can ever be attained, given the obscure situation of the source material. Future work should include a close reading of both texts and possibly an exact translation of the Latin Apologia, so that it can more easily be compared to its English version. This seems especially fruitful for the first part of the treatise since parts two and three do not contain any arguments, just testimonies and medicinal use cases. It seems unlikely at this point that stylometric analysis could shed further light on the matter at hand. Yet it remains a desideratum to re-run the analyses on manually cleaned and thus, less erroneous transcriptions. The results of these analyses are a good example of the great difficulty in evaluating stylometric results, specifically how strong an attribution is and what it really means in terms of authorial involvement. [2] F. Anthony. Medicinae chymicae et veri potabilis auri Assertio in Francisci Antonii philosophi et medici Londinensis panacea aurea sive tractatus duo de ipsius auro potabili.
Cambridge: C. Legge, 1610. [3] F. Anthony. The Apologie, or Defence of a Verity Heretofore Published Concerning a
Medicine Called Aurum Potabile. London: John Legatt, 1616.
W. Beck. Michael Maiers Examen fucorum Pseudo-Chymicorum – Eine Schrift wider die falschen Alchemisten. München: Diss., TU München., 1991.
M. Maier. Arcana Arcanissima, hoc est, Hieroglyphica Aegyptio-Graeca, vulgo necdum cognita, ad demonstrandam falsorum apud antiquos deorum, dearum, heroum, animatium et institutorum, pro sacris receptorum, originem, ex uno Aegyptiorum articio, quod aurem animi et Corporis medicamentum peregit, deductam, Unde tot poetarum allegoriae, scriptorum narrationes fabulosae et pertotam Encyclopaediam errores sparsi clarissima veritatis luce manifestantur, suaeque tribui singula restituuntur, sex libris exposita. London, 1614.
M. Maier. Civitas Corporis Humani, a Tyrannide Arthritica vindicata: Hoc est, Podagrae, Chiragrae, et Gonagrae, quae velut tyranni immanissimi artus extremos obsident, et excruciant, Methodica Curatio. Duobus auxiliis potissimum instituta, ac deinde latius clarissimorum, praesertim GERMANIAE, Medicorum testimoniis comprobata, inque Medicinae Candidatorum gratiam atque utilitatem concinnata et edita. Frankfurt am Main, 1621.
M. Maier. De Circulo Physico, Quadrato: Hoc est, AURO, Eiusque virtute medicinali, sub duro cortice instar nuclei latente; An et qualis inde petenda sit, Tractatus haud inutilis. Oppenheim, 1616.
M. Maier. De medicina regia et verè heroica, Coelidonia. Prag, 1609.
M. Maier. Examen Fucorum Pseudo-chymicorum detectorum et in gratiam veritatis amantium succincte refutatorum. Frankfurt, 1617.
M. Maier. Jocus Severus, hoc est, Tribunal aequum, quo noctua regina avium, Phoenice arbitro, post varias disceptationes et querelas volucrum eam infestantium pronunciatur. Frankfurt am Main, 1617.
M. Maier. Septimana Philosophica, qua aenigmata aureola de omni naturae genere a Salomone Israelitarum sapientissimo rege, et Arabiae regina Saba, nec non Hyramo, Tyri principe, sibi invicem in modum colloquii proponuntur et enodantur: ubi passim novae, at verae, cum ratione et experientia convenientes, rerum naturalium causae exponuntur et demonstrantur, figuris cupro incisis singulis diebus adjectis . Frankfurt, 1620. M. Maier. Symbola Aureae Mensae Duodecim Nationum, Hoc est, Hermaea seu Mercurii Festa ab Heroibus duodenis selectis, artis chymicae usu, sapientia et authoritate paribus celebrata, ad Pyropolynicem seu Adversarium illum tot annis iactabundum, virgini Chemiae Iniuriam argumentis tam vitiosis, quam conuitiis argutis inferentem, confundendum et exarmandum, Artifices vero optime de ea meritos suo honori et famae restituendum. Ubi et artis continuatio et veritas invicta 36 rationibus, et experientia librisque authorum plus quam trecentis demonstrantur. Opus, ut Chemiae, sic omnibus antiquitatis et rerum scitu dignissimarum percupidis, utilissimum, 12 libris explicatum et traditum, figuris cupro incisis passim adjectis. (Nachdruck: Akademische Druck und Verlags Anstalt, Graz 1972.) Frankfurt, 1617.
M. Maier. Tripus Aureus, hoc est, Tres tractatus chymici selectissimi, nempe I. Basili Valentini, Benedictini ordinis monachi, Germani. Practica una cum 12. clavibus et appendice, ex scriptum, nunc ex Anglicano manuscripto in Latinum translatum, phrasi cuiusque authoris ut et sententia retenta; III. Cremeri cuius Abbatis Westmonasteriensis Angli Testamentum, hactenus nondum publicatum, nunc in diversarum nationum gratiam editi, et figuris cupro afabre incisis ornati opera et studio Michaelis Maieri. Phil. et Med. D. Com. P. etc. Frankfurt, 1618.
M. Maier. Viatorium, hoc est, De Montibus Planetarum septem seu netallorum; Tractatus tam utilis, quam perspicuus, quo, ut Indice Mercuriali in triviis, vel Ariadneo filo in Labyrintho, seu Cynosura in oceano chymicorum errorum immenso, quilibet rationalis, veritatis amans, ad illum, qui in montinus sese abdidit De Rubeapetra Alexicacum, omnibus medicis desideratum, investigandum, uti poterit. Oppenheim, 1618.