Aristotle ' s Metaphysics : why should Harlfinger ' s stemma be verified ?

The textual transmission of Aristotle’s Metaphysics is currently described by Dieter Harlfinger’s stemma codicum. It appeared in 1979 within the acts of the 1972 Symposium Aristotelicum. 1 With a single exception, the stemma has been accepted by scholars without discussion, or with minor relevances only. On the other side, at least until 2009 no stemmaticallybased edition of a single book of the Metaphysics appeared. Still today, no new general edition is available. We are thus still left with Jaeger’s 1957 OCT – admittedly, an editio minor, which partly depends on Ross’ 1924 critical apparatus and textual choices. But things are evolving now, as we are about to see: this crucial theory and practice – editing Aristotle’s Metaphysics –is moving today faster than it has since the 19th century. Hence the interest in promoting a broader and a more articulated discussion, by pointing out some basic desiderata, which show the need for the subject to be taken into consideration anew. But since the field has been asleep for a long while, I’ll begin, in section 1, with some preliminary remarks. There I start with the most general question, about what our sought-for edition is expected to be (1.1), and more exactly why a stemma codicum, if available, can be so helpful for this purpose (1.1.1); then, I outline the present status of the art and its evolution during the 20th century (1.2), with special reference to Bernardinello’s and Harlfinger’s works, and to their connections, on which I will also report (1.3). All this is intended to make my points and doubts in the main sections 2 and 3 more perspicuous in the frame of their historical and methodological background; however, readers who are already familiar with the matter at issue are welcome to skip preliminary sections 1.1 to 1.3, proceeding to my main discussions in sections 2ff.


Ways of conceiving the best possible critical edition
Let me then start again with a very general question.What is the best possible critical edition of Aristotle's Metaphysics one can dream of?If the answer is: 'an edition one can actually make sense of, and can actually read', we are, so to say, in the mainstream.This is the way current editions in modern languages are mostly intended to work.In many regards, nonetheless, it can be argued that such a criterion does not satisfy scholarly purposes.For it leaves unexplained whether such a sought-for sense is really the one intended by the author.And even if one further specifies, e.g., that 'the best edition is such that one can make the right sense of it", this is either likely to put an editor's task out of the range of reasonable targets (Aristotle's intention: who knows?), or to make it, on the contrary, exceedingly handy for most compelling interpretations.In a sense, this is the way Jaeger 's  edition  (1957) has been intended to work.Not much of Ross's critical apparatus has been improved (especially since Ross, on his turn, had taken into account most of Jaeger's early remarks and suggestions).Some textual notes get even worse and more confused, as a kind of second hand work. 3Even Ross's textual choices remain for the most part unchanged.In a minority of the cases only, Jaeger's text is crucially different from Ross's.There, leaving aside the cases where their different judgements concern nothing but the style of textual minutiae, Jaeger seem to have seen his task as producing the text most suitable for his own understanding of it.This also includes, as a peculiar general feature, Jaeger's famous Doppelklammern (with the form "[[…]]"), which he uses to set apart, but not to reject, bits of Aristotelian text, with the idea that those "Aristotelis additamenta", as he calls them, were introduced into the main text by Aristotle himself (in principle; or by his personal assistants) thus making, with their presence, the difference between our actual text and the Ur-Metaphysik, as Jaeger calls it. 4In fact, in a book like Lambda, where Jaeger's 1957 edition has Doppelklammern just three times, it is not obvious what the rationale for such an insertion on Aristotle's part is supposed to be. 5 Jaeger does not say.Perhaps, as it appears, it is the very fact of finding at least some Doppelklammern, which is meant either to confirm Jaeger's overall genetic thesis (with some risk of circularity in that case) or at least to strengthen the continuity in thought within the forty-five years of Jaeger's intellectual career.In general, what seems to be the leading criterion of Jaeger's editorial choice is his own (as an editor's) iudicium, as he calls it in his 'Praefatio', where he summarizes his theory about it.With this (as obvious as it may seem), Jaeger reacts to the opposite tendency, the one reducing an editor's intervention to a ranking of readings, based on codicum conspiratio, as he dismissingly says in his "Praefatio" (xv f.), with obvious reference to stemmatics.This brings us back to our starting question: what the best possible critical edition of Aristotle's Metaphysics is supposed to be.

Striving for a textual archetype: the stemmatic method
In principle, a rational method exists, thus making the creation of an edition as uncontroversial as possible, led by rational criteria.The method can be spelt out and reached by means of stemmatology, the scientific, genealogical approach to a text's variants readings.This implies mutual relationships within all extant manuscripts to be represented as a stemma, i.e. as a family tree, showing which one is an apographon (a "child") or an antigraphon (a "parent"), or a nephew or an uncle of another, including those manuscripts which are now lost (these latter being usually labelled with Greek letters).Parental relationship are basically detected through agreement in errors, with special reference to significant errors, e.g.textual lacunae, which are unlikely to be spontaneously detected and corrected within a copying process, or reproduced identically in more manuscripts if one is not the copy of the other.The idea is that every copy has all the (significant) errors of the model plus further errors, thus allowing modern scholars to trace back their mutual, genealogical relationship.Unfortunately, the procedure does not work if scribes have used more than a single modelas they do in particular when they aim to improve the received text while checking one manuscript against the other.For in this way, they correct errors and lacunae.In a word, they do not merely act as copyists but rather as editors, so that there is no mechanical way to discover the path they followed.From a lachmanian, i.e., stemmatic point of view, there is a 'contamination' from a plurality of sources.(This kind of phenomenon, according to Harlfinger, affects the tradition of the Metaphysics as well, but not the oldest manuscripts, as we are about to see).
In those cases, where all manuscripts have a single ancestor or archetype, a reconstruction can be achieved, errors can be detected as such, and an editor's target can be reached, as in Maas' definition: 'The business of textual criticism is to produce a text as close as possible to the original (constitutio textus), […]  known as the archetype '. 6 This way, based on factual data, if the stemma is established well enough, an edition can be reached in a rational way such that subjective editorial 6 P. Maas Textual criticism.Oxford, 1958, 2f.I will not deal in my present note with the different steps before this precise target, or out of it, as for example those which are a matter, either of philological conjecture (emendatio, especially for correcting errors), or of historical speculation based on external hints.choices are reduced to a minimum. 7Things are much more controversial if there is not one but two (or more) reconstructible archetypes (recensione aperta).Still, this can be regarded as a second best case, provided that the individual archetypes (which some prefer to call "hypo-archetypes") can be properly reconstructed.It depends then on the editor and on any interested scholar's iudicium to choose, based on the critical apparatus, between the variant readings of the thus-reconstructed master copies (Vorlagen).Standard criteria are then in use (see below), like the usus scribendi (the knowledge of an author's style) and the lectio difficilior (a kind of subtle but statistically valid approach, based on the likelihood of more difficult readings to be dropped out from the text, as opposed to simple and more expected ones, which are more easily introduced instead).
However, in order to detect the overall structure of the stemma, nothing sound can be done before an accurate eliminatio codicum descriptorum.This cannot but be a heavy and timeconsuming task.Its importance is manifest.By contrast, let's suppose that an editorially-elaborate branch of the transmission, made of copies and copies of copies of still extant exemplars, is misunderstood as the trace of an ancient ancestor from late antiquity; there will be no obstacle to its smoothened readings to be preferred i.e. inserted into the main text of a critical edition.
Hence the importance of Silvio Bernardinello's eliminatio, which led to the first stemma codicum of the Metaphysics, and will be described in § 2.1 below.8

The present status of the art of the Metaphysics' Textüberlieferung
Due to a large consensus in 19 th and 20 th scholarship, the textual transmission of the Metaphysics is expected to be a recensione aperta, thus an instance of the latter case seen above.This idea virtually dates from Brandis edition (1823), when readings were firstly introduced from the Florence manuscript Ab (Laur.87.12; for all of these sigla see our "Appendix 2" here below).
Readings in Ab are slightly different in style and wording from any other manuscript in use until then, all of which belong to the alpha family.Belonging also to this family are Bessarion's manuscripts, whose text, throughout revised by Bessarion himself, was indirectly the source of Metaphysics in Aldus Manutius' 1497 editio princeps of the Aristotelian corpus (see § 3.2.3below).
Aldus' text in its turn was slightly revised for the sake of further editions, and was somehow probably the basis of Erasmus' and of all subsequent editions until Brandis'.9 The relevance of Ab, a different kind of manuscript from those hitherto circulated, makes now the text of the Metaphysics out of two distinct recensiones, which were described in detail by Christ, Ross, and Jaeger.Each of these scholars claims to have collated manuscript Ab better than previous editors dida claim which shows by itself the importance paid to the Florence manuscript, regarded as an authorized representative of a different tradition (recensio).The credit of Ab among editors was then taken in due account, almost as a proof of its independent value, once the first overall stemma of the Metaphysics was drawn by Silvio Bernardinello (1970).
Before this, neither Ross nor Jaeger prepared or used a stemma.10Among several reasons, one has just been mentioned: the stemma codicum presupposes an eliminatio codicum descriptorum.
This in its turn presupposes a comprehensive list of all extant manuscripts to be available.In fact, until 1957 there was none -either of the Metaphysics or of any other Aristotelian text.Not that manuscripts were not listed yet in library catalogues, but only in 1948 was a general repertoire of Library catalogues collecting ancient manuscripts first printed.11On this basis, a catalogue of all Aristotelian manuscripts was accomplished by André Wartelle, on Paul Moraux, Jean Irigoin and Alphonse Dain 's demand, in 1957. 12 Based on more than 500 catalogues, Wartelle made thus available an overall list of more than fifty manuscripts of the Metaphysics (an English version, with Harlfinger's description, is appended here below).On this very basis, Silvio Bernardinello undertook his "Eliminatio codicum della Metafisica di Aristotele".When in 1970 he published a monograph with this title, devoted to this task, this was the first step toward a stemmatic approach to our text.
Strictly speaking, Bernardinello's eliminatio is a selection: the editor sets apart the codices descripti, i.e. those manuscripts which are just copies of other exemplars we have, so that they should not be taken into consideration at all for the sake of a critical edition.
2.1.The Metaphysics' stemma codicum: a comparison between the first proposal (Bernardinello,   1970) and the current standard (Harlfinger, 1972/79)   The first attempt of a Metaphysics' stemma codicum was then proposed by Silvio Bernardinello in 1970 within his crucial eliminatio codicum descriptorum.Bernardinello, as we mentioned, assumed as a likely starting point that manuscript Ab is independent from E and J, i.e. independent from the most ancient manuscripts of the Metaphysics, and from their common source.
Confirming this basic assumption, Bernardinello found the most significant support available to us, a fragment from the 10th century, which he labelled Y (Paris.Suppl.gr.687).He firstly pointed out that Ab and Y coincide in most of their significant readings14 .Bernardinello first called J's and E's source alpha, as opposed to beta.The latter family could be thereafter seen as a source not only for Ab, as it could have seemed until then, but for Y and for several other manuscripts.The relation between the two proposals consists both of difference and of agreement: it can be described in terms of difference within an overall frame of agreement, which is given as obvious in Harlfinger's context, but deserves nowadays to be briefly recalled, especially since Bernardinello presents data which Harlfinger does not, whereas Harlfinger spells out some of his argument especially when he has to emphasize the difference of his own findings.
The harshest disagreement is spelled out by Harlfinger in three points, none of which affects either the textual constitution, nor the reconstruction of the source for the history of printed editions.
Such remarks are mainly intended to show, as Harlfinger introduces them, that the difference between the two stemmata cannot be neglected. 15Here Harlfinger's perspective and willingness to emphasize differences seem to play a role in the way the cases at stake are discussed.In all of the three cases, once they are seen from the outside, agreement could seem to be more relevant than disagreement.In all of them, even if Harlfinger were right in all of his points of criticism, Bernardinello deserves credit for having detected a number of tight relations between the three relevant couples or groups of manuscripts (S, s, and t; Ja and Dm, Pa and Bb), all of which are confirmed in Harlfinger's stemma.Where Pa and Bb are concerned, Harlfinger tacitly agrees that the two are closely related, as he makes them brothers under a deperditus without name, rather than father and son, because of "Trennfehlern in Bb" which unfortunately he does not spell out nor locate in the text.When S, s, t are concerned, Harlfinger agrees that t is a copy of S, but he reverses the role of model and copy between s and S, while strongly disqualifying Bernardinello's dating of S as impossible because of the paper and the ductus.The fact apart that all of the contested points are given very cautiously by Bernardinello, this latter already gives some counterarguments in Harlfinger's direction (within his twenty pages discussion around the composite manuscript S): he does firstly point out that S's paper seems to be old; as for S's ductus, he discusses as unexpected Vogel-Gradthausen's 15th century dating of the 4th copyist at work on S, apparently the same for De caelo and Metaphysics, vs. the Italian catalogue by Bandini, which Bernardinello quotes, which already has S at the 13th century as Harlfinger has.By contrast, Harlfinger's denial does not enter into Bernardinello's disputed questions, e.g.Harlfinger does not say whether he finds that Vogel-Gradthausen are wrong too, or that the Metaphysics is not by the same copyist as De caelo is; nor does he say why, if S is the source of s and not vice versa, S omits physeôs at 986b11 whereas s does not (this short lacuna being the reason for Bernardinello 's  final view, p. 200).We are simply asked to trust Harlfinger's judgement.
And when Harlfinger says that Dm, though being connected with Ja, cannot be one of the copies of Ja, he still tacitly allows that Nd, Vc, Mc, d, Qc are so, and that O in its turn is a copy of Nd, and Q of Dmthese all being genuine findings made by Bernardinello.
At any rate, none of the points of detail Harlfinger directly makes against Bernardinello, is relevant to the textual constitution.On the whole, as I suggest, these points are probably less worth considering than some crucial overall features of the new stemma Harlfinger proposes, which I 15 "Was darüber hinaus das Stemma codicum anlangt -Harlfinger writes, p. 8weichen die hier vorgelegten Ergebnisse von denen Bernardinellos nicht unbeträchtlich ab.Es würde zweifellos den Rahmen dieses Beitrags sprengen, wenn bei der Vielzahl der anliegenden Probleme die Auseinandersetzung mit den Resultaten Bernardinellos im einzelnen durchgeführt würde.Statt einer Diskussion seiner Ergebnisse im jeweiligen Zusammenhang seien hier nur drei Beispiele angeführt, die ohne ausführliche Erklärungeneinleuchten." would regard as the main difference between the two.I refer namely to the thorough contamination which crosses the stemma and affects the status of every single manuscript or branch of manuscripts except the vetustissimi J, E, Y and a couple of supposed copies of them, such as T and X.
The two families are shown to be both at work in manuscript Ab too, albeit in a peculiar way: manuscript Ab would switch at some point from a family, beta, to the other, alpha.More exactly, following in Bernardinello's footsteps (although this latter is not mentioned on this regard), Harlfinger has shown that at least starting from book Lambda, ch. 8, Ab is not independent from alpha, the common source of E and J: it is an alpha manuscript 17 .Harlfinger argued moreover that more or less all of Bernardinello's remaining beta manuscripts which are complete and independent share common readings with the alpha tradition, a phenomenon which he explains by saying that most of them belong to the alpha family, though they are affected by a large contamination from the beta family.As for the few remaining ones M, C, Vk, they still are beta manuscripts in Harlfinger's view as well, but not without alpha contamination.As a result, all of them happen to have more than one source, so that the weight of one branch against another is very hard to establish in detail: "gegen die Kontamination ist kein Kraut gewachsen" (P.Maas, Textkritik, Teubner, Leipzig 1957, p. 31).
On the other side, always in Harlfinger's view, all of alpha independent manuscripts too, except E and J, bear contaminations, namely from beta manuscripts.This is seen in Dieter Harlfinger's stemma, the standard since 1979 (see n. 1 above): 17 Bernardinello, op.cit., p. p. 136, corsivi miei): «La parte, nella quale le lacune dei due codici [Ab and M] non concordano più, corrisponde quasi esattamente all'intervento del copista più recente in Ab ».The most recent copyist, in Bernardinello's view as in Harlfinger's, is at work from f. 485v, i.e.Lambda 1073a1, so the two scholars seem to agree on this point too.See however my Lo stemma codicum dei libri Kappa e Lambda della Metafisica: una revisione necessaria, «Aevum», LXXXIV (2010) pp.339-59, where I show that already by 1065a (Kappa 8) Ab cannot be proved to have an independent source from alpha, so that its reading has to be regarded, thereafter at least, as a partly revised, partly corrupted version of the alpha text.If my suggestion is correct, and if the major affiliation change in Ab is not at 1073a1, we would have a further agreement in error among the two scholars.However, Bernardinello's position is stated more cautiously.
Harlfinger's stemma remained entirely undisputed until 2010.Perhaps, it was regarded as undisputable far beyond the author's intention, who delivered it as a draft rather than as a final version, and acknowledges more than once that he is relying on data he does not publish. 18  1 See e.g.Harlfinger's references to "other collations of his" at p. 12, 14.
However, the status of the art of editing the Metaphysics is awake today, and moving in more than one direction.Times are more favourable now.Primavesi's Alpha edition was then published in 2012 within the acts of the 2008 Symposium Aristotelicum devoted to this book.Indeed, in spite of their different geneses, these two 2012 editions have in common some well-connected features.Firstly, both followed Dieter Harlfinger 's   (1979) stemma codicum, which is now regarded as a standard; a major revision was suggested for the Kappa 8 to Lambda 8 section only, concerning ms.Ab: this was affiliated to alpha and not to beta as it was by Harlfinger. 21In both editions, according to the reference stemma, not every independent manuscript was quoted in the apparatus (unlike what Bernardinello 1970 had suggested, and Stefan Alexandru did) but a tight selection of them only (broader in Primavesi, though): manuscripts in use have been thus severely restricted to those which are needed for reconstruction of the two stemmatic archetypes or to make sense of former editor's textual choices (which may justify the collation of Ab's readings in the apparatus even when this manuscript is nothing but a recentior, a comparatively unreliable witness of the alpha family).This implies a step forward in rationality, a progress which deserves to be emphasized.The editor is thus left with the pure alternative between alpha and beta.On this basis, different readings (variae lectiones) are to be ranked and selected.Primavesi's edition of book Alpha is a prominent example of a mature Tradizione, PhDTh, Université de Lille-Nord de France and Università di Trento, 2009.This latter splits in three parts, namely, the critical edition, a commentary and a collection of contributions on Lambda's reception: this was then the first stemmatically-based critical edition of a book of the Metaphysics, although it implied a relevant revision of Harlfinger's stemma codicum, one concerning ms.Ab (published a part in 2010, see the following note). 21Fazzo "Lo stemma codicum dei libri Kappa e Lambda della Metafisica", cit.. reflexion on this kind of open 'recensio'.It presupposes that the two archetypes are comparable in value.As Primavesi says (p. 409), "it is just stating the obvious to say that not only scribal errors but also intentional modifications of the original text may have occurred on either side".If so, it remains to an editor's iudicium to decide between the two branches, based on such criteria as the lectio difficilior and the usus scribendi.Moreover, for books Alpha-Delta (Alpha included), unlike Epsilon-Nu (Lambda included) Alexander of Aphrodisias' commentary is preserved, where some minutiae seem to be in agreement with Ab, and M.These points of agreement are already put to use by Jaeger, Ross, and Primavesi, where they play a role especially in cases of doubt (in Hermann Diels' footsteps22 ).
Primavesi, although he generally prefers the alpha manuscripts, 23 follows Ab in cutting away some nine or ten passages which are in the main tradition but are not found in this latter manuscript. 24He has somehow the stemma on his sideat least so far as we can judge from the stemma 25 .
Hence, once more, the importance of the distinction between the two families to be surely established and the reference stemma to be verified in depth.
As an alternative, by the way, in 2013 a proposal has been made to enhance the importance of sources whose stemmatic position has not been established yet, such as the variant reading of recent manuscripts, or pieces of indirect evidence, such as translations in different languages and second hand variants and glosses. 26Recent manuscripts, however, raise a crucial problem because none of them, according to Harlfinger, escaped contamination.It is not certain, therefore, that their position can be ascertained at all.Such a tendency, if concretely developed, would favour a decrease in rationality and decidability of those editorial options, which up to now could be performed on a stemmatic basis.If such a trend is to be established, no consensus on a standard reference edition of the Metaphysics will be reached in the near future.Willingly or not, this kind of proposal raises a question to be answered by future developments in this field: how can one contribute to producing a more accurate, i.e. (as I take it) a more reliable critical text of the Metaphysics, as opposed to a larger, and more confused, critical apparatus (as if one publishes the readings of every source whose origin and authority has not been proved and properly classified yet).
As a matter of fact, the proposal to make use of later manuscripts and translations is related to several zones of opacity which seems to affect the status of the art of text criticism and call for clarification.These are mainly connectedas I see it -to the lack of discussion around Harlfinger's stemma, in spite of the credit it has received.

Open desiderata around the stemma: dressing a list, from top to bottom, from left to right
Hence the interest of pointing out now some desiderata and some shadowy zones, whose exploration may help to strengthenand at the same time, to revise and update -Harlfinger's stemma.
Shadowy zones are to be found more or less everywhere in the stemma, due to the subtlety of the subject matter.The most important of them concern the upper part of the stemma, i.e. those which we may call our codices vetustissimi.By vetustissimi, I mean especially J, from the 9th century, which has two hands on it, and E, from the 10th 28 ; moreover, the Y beta fragment, discovered by Bernardinello, which is also from the first part of the 10 th century, and has never been collated in full.On the other hand, if we proceed left to right, we find obscurity zones both in the left-hand part of the stemma, i.e. the so-called alpha family; and even more, in the right-hand part, the beta family.Some are more crucial than others.I will try to list the most interesting of them.
The most important task for scholars to come is probably a deeper inquiry on the top part of the stemma.One has to clarify or at least, to describe step-by-step the relation between those few witnesses, J1, J2 (see Appendix 1 here below), E, and possibly Y, all of which must be closely connected either with the first phase of minuscule transcription or with the first phase of editing procedures.It would be relevant to check their relationship and their location in the given contexts.
One has to see, for example, which kind of variant readings make the difference among them; 28 As it appears, see Appendix 1 below, J is made from at least two hands, a first scribe and a diorthotês, both in the 9 th century, i.e.J 1 plus J 2 .As for now, this can be assumed just by comparing the two styles of handwriting, e.g. in the way nexus are drawn, or a same word is written.But the matter needs further exploration, especially since they have been recently held (see the previous notes) to be one and the same.moreover, one has to detect differences and similarities in their way of being written onto the page, their punctuation, their word division-in a word, everything that could have been easily transmitted from one copy to another.
As for the bottom, most recent part of the stemma, most of it is not relevant for the textual constitution, as if it were, from an editor's point of view comburendi, non conferendi, (according to Cobet's motto, as adopted by Paul Maas, "Ruckblick 1956" in Textkritik, Leipzig 1957 3 , p. 31ff., to evaluate such recentiores).Still, a few of them appear to bear some special interest for textual history: as mentioned above, it is worth considering what source was used for the printed texts of the Metaphysics, starting from Aldus' editio princeps in 1497 (not 1498, as it is usually written) until Brandis in 1823 (the first modern editor undertaking new collations). 29For in this regard some crucial progress has been made after the first draft of Harlfinger's stemma, and the final draft does not take it into account, although the relevant information about the role of ms.Qc is given as note "56bis".So I'll recall the point and give some further detail (see § 3.2.3below).
As for the alpha and beta (hypo)-archetypes, this is where the main and most difficult desideratum is: one has to clarify everything one can grasp, based as far as possible on extant documentary data, i.e. by relying as little as possible on speculative reconstruction.Then, as a second priority, the one concerning further number of hypo-hypo-archetypes, with special reference to the highest alpha hypo-archetype, namely gamma, and to both beta hypo-archetypes: epsilon and zeta.This priority is due to the fact that their establishment (as opposed to others') is likely to affect both the critical apparatus and the selection of readings.I start with epsilon and zeta.A common feature can be found not in the errors they (and their entire subfamily) contain, but in the errors they do not contain.Nonetheless, every one of them must have injected into its apographoi some peculiar errors of its own, some of which go against (versus) all of the remaining manuscript tradition.In both cases, such an inquiry about separative/conjunctive errors needs to be undertaken, either entirely, as in zetafor which no Trennfehler is given by Harlfinger, or in part, as in epsilon a deperditus for which only Trennfehler within book Kappa are given by Harlfinger.More specifically, no peculiar series of errors can be detected in book Alpha meizon (judging at least from Primavesi's published collations), nor in Eta-Theta (Harlfinger's collations).
As for hypo-archetype gamma, by contrast, and even for J alone, Harlfinger does spell out a series of peculiar readings.The question is how can one establish that such readings are errors?In a sense, a stemma would be needed in order to judge on this matter; one understands Bernardinello 's 29 Brandis is also the first editor who introduced ms.Ab into useas Albert Schwegler tells us, although he does not spell out what his manuscript sources were: he has a critical apparatus with variant readings only but without manuscripts' sigla.
choice, to establish his stemma based on lacunae only, namely on the most obvious kind of errorsif there's one. 30re generally, one now expects all of Harlfinger's assumptions to be verified based on those full fresh collations of all the manuscripts of the Metaphysics (maybe a new elimination codicum would not be superfluous, since Bernardinello's was so harshly criticized), or to be revised accordingly.
Perhaps, the entire view we have of the beta family is now collapsing.Admittedly, though, a new scenario to be reconstructed is not going to be an easy task.

Looking at the central and right hand bottom zone of Harflinger's stemma
Most of all, as argued so far, one has to explore all of Harlfinger's proposed hypo-archetypes: zeta, epsilon, gamma, beta.This whole enquiry is a demanding task, on which I hope to say something more in the future.
For now I append a more detailed discussion of issues related to a lower, less crucial and particular zone, to which Bernardinello and Harlfinger seem to have paid considerable attention, because of the high number of manuscripts it includes: the bottom of the left-hand part of the stemma, represented by Harlfinger thus: 30 See however below.
The fact that C lacks not less than ten Bekker pages, i.e., almost the entire book Gamma plus the final parts of Beta and the beginning of Delta, seems to have escaped attention, especially since such large lacunae are not at the beginning (or at the end), but later on.More exactly, the text is missing from Beta 1001b12 (end of the preliminary section without regular folios numbering) to Gamma 1006b12 (beginning of three extant Gamma folios, ff.4r-v, 3r-v, 2r-v, see below), and then is missing again from Gamma 1008b17 (f.5r) to Delta 1013a18.In sum, one could also say that that there is a single major lacuna from Beta 1001b12 to Delta 1013a18, with just a few fragments from Gamma 4, 1006b13-1008b17, in between.These fragments, being in the bad condition as Bernardinello describes them have not been recognized well enough to allow their right location: in the actual state of the repaired manuscript, the position of the three folios at issue is inverted: f. 2r-v has, although in part only, Gamma 1008a7-b17, f. 3r-v has 1007a28-b31, f. 4r-v has 1006b12-1007a28.[See figure 2 below, with the first readable folium of the main section, which has been

C and Jc as father and son in book Alpha?
Given that the status of C is unclear, how can Harlfinger claim that its text of book Alpha (for the remaining books, see below) is the antigraphon of a manuscript, Jc, which is found to be an apographon of a printed book, namely of Aldus' edition? 36The point is relevant for Primavesi's new critical edition of book Alpha (see above).In his introduction Primavesi says ms. C is partly damaged by the burning, so it is worth replacing it with its apographon (p.395 and ibid.fig. 3 and n. 36, p. 398 and ibid.fig.5).Then he often repeats Harlfinger's assessment, calling Jc an apographon of C.He never says the extent to which C is burned, but in context he must be referring to the marginal partial burning of book Alpha, since he gives this as a reason to introduce in the critical apparatus C's supposed apographon, i.e., Jc, a manuscript which Bernardinello strongly suggest not to take in a critical apparatus. 37Yet, now Primavesi's published collations (or Golitsis' under Primavesi's direction, see his p. 387 and n. 12 above) seem to show (so far as I can see) that Jc is not an apographon of C. In particular, at A 4. 985a19-20 a large lacuna is in C, and in AbMVk, but not in Jc, and this should be enough to show Jc not to be an apographon of C; dozens of other times,moreover, Jc does not agree with C, and in many of them Jc joins other manuscripts against C. 38 As for me, I would be sure and confident that Harlfinger gave his assessment with due and precise reason, but something must have gone wrong at some point: what did he and Bernardinello intend to say?As for now, the point has some obscurity and can be taken as an example of the kind of issue which calls for verification. 36See Harlfinger, art.cit., p. 28: "Für Buch A ist Jc nicht Druckapographon wie für die Bücher a K N, sondern der Kopist Matthusalas Macheir schrieb es kurz vor 1562 wahrscheinlich in Konstantinopel aus dem Kodex C des Kamariotes ab.".No argument is given, nor it is clear what the relevance and reason for the adverb "wahrscheinlich" is. 37See Bernardinello, op.cit., about Jc and its apographon Jd: "entrambi i codici sembrerebbero un falso commerciale di Mathusalas ai danni del Busbeck" (p.82);; "Pertanto questi codici e l'Aldina non dovranno comparire in un apparato critico che sia veramente degno di tal nome."(p.88). 38See Primavesi 's  critical  apparatus  at  A 2. 982b5, A 4. 985a19-20, 22, 23, 27, 985b15 (twice), 17, 19, A 5.  985b24, 25, 986a4, 11, b3-4, 9, 17, A 6. 987b11, and so on.See also in the last chapters, cf.A 8. 990a25, 25-26, 28 (twice), A 9. 990b5, 7-8, 8-9, 12, 18-9, 32, 34 (where it is only a difference in diacritics), 991a4, 9, 10,  23, 29, 991b1, 11, 25, 28, 30.992a1, 6, 8, 11, 13, 15, 16, 20, 24, 26, 29-30, 992b6, 9, 12, 12-13, 14, 15, 16,  17-18, 18, 29, 993a1, 2-3, 5 (twice), 8, 12, 16 (twice), 19-20, 24, 27, and so on.Metaphysics  (1947, "a"  in the stemma)?

What's the source of Aldus Manutius' edition princeps of the
The last example I will give concerns a point which does not concern the text's Bernardinello ranks the Aldina among those textual witness which originate from Ja, a manuscript closely connected to Bessarion, together with Qc and many other manuscripts: eight of them directly, Qc included, nine indirectly, among which two copies, Jc, Vind.Phil.189, and Jd, Vind.Phil.217, of Aldus Manutius' printed text (see the below detail of Bernardinello's stemma about the succession a > Jc> Jd).So on the whole more than a third of all extant manuscripts of the Metaphysics are connected with Bessarion's copies and textual revisions. 4039 The explicit of the relevant Aldus' 4.2 volume says: "Excriptum Venetiis in domo Aldi Manutii Calēdis Iunii .M.III.D".Aristotle's Metaphysics, with Theophrastus', is at the end of volume 4.2, in Aldus 5 volumes 1495-1498 edition.The 1498 volume, which is the last one, contains ethical and political works of Aristotle: Nichomaean Ethics, Politics, Economics, Magna Moralia, Eudemian Ethics.This implies among other things, an arrangement of the Metaphysics within the corpus after the pseudo-Aristotelian Mechanics and before the ethical and political works which is partly reproduced in Bekker's standard edition. 40This makes particularly impressive, and not just for the Metaphysics, the role of this Orthodox priest converted and made Cardinal Bishop as a Roman Catholic.Bessarion's activity as a collector and reviser of manuscripts was so influential because he left all of his manuscripts to Venice's Republic, under request of putting them in the middle: he successfully asked a public library to be opened for the sake of anyone wishing to access the manuscripts, so that this is why the Biblioteca Marciana was founded in Piazza S. Marco in Venice.Bessarion's editing activity was crucial for the edition princeps of other works too, eg Alexander's of Aphrodisias' independent works (Opera minora, as the CAG edition labels them).The probable archetype of a part of them is preserved in the Marciana Library as Ven.258, with glosses and corrections by Bessarion himself which were then included in the first printed text of the work, Venice 1538, and partly in the CAG edition as well (ed.by Ivo Bruns as CAG Suppl Ar.II.1-2, 1889-1892).
are argued to bear the trace of one or more authorized codices deperditi.This is why, as Giorgio Pasquali taught, recentiores non (ie, are not necessarily) deteriores.That is: recent manuscripts should not be disregarded from the outset based on their age.Harlfinger's stemma follows this advice, when pointing out the value of C and M as beta witnesses, when it appears Ab cannot be trusted.With this in mind, Harlfinger too has Pasquali's motto as an explicit of his brilliant stemmatological performance, "Zur Überlieferungsgeschichte der 'Metaphysik'".
In the last decades, editors of the Metaphysics are in a better condition than ever because of Harlfinger's current stemma.This is highly superior to any previous editing tool, even if Bernardinello's eliminatio codicum descriptorum and first draft of a stemma deserves substantial credit and attention as a preparatory work which is not entirely superseded, as I have noticed so far.
The strength of Harlfinger's stemma, as I suggest, is not so much in some emphasized differences with Bernardinello, but in the way it manages to put into value subtle and sometimes crucial internal relationships within codices and groups of codices which escaped attention before, and could eventually allow us to eliminate even more among the relevant codices for a critical edition.
This can be especially seen with reference to Harlfinger's system of hypo-archetypes.It is relevant for textual choices to establish whether such hypothetical deperditi as gamma (within the alpha family) and epsilon (within the beta family) existed or not.They mean that concordances in particular readings within their sub-families do not testify to anything but common errors, whereas concordances in disagreement from those particular readings, between manuscripts which do not belong to those particular sub-families, are shown to preserve the original wording, from before such hypo-archetypal errors was introduced.In such a context, it would be crucial to detect, whenever possible, and to submit to open and careful discussion where the correct text is to be found for each case.This is not always possible, because many concurrent variae lectiones are neither manifestly wrong nor right: their judgement requires the stemma to be firmly established. 42nce, in Bernardinello's and in Ross' view, the stemmatic importance of such patent errors like lacunae. 43Still, lacunae alone are not enough to allow a clear draft of the stemma, as shown by Harlfinger, which appears to take into account a large amount of further data. 42See for a proposal of statistic likelihood according to the type of readings, Marwan Rashed's introduction to Aristotle: De la génération et de la corruption, Les Belles Lettres, Paris 2005, CLXXXVII-CCLII, in particular CCXXIV-CCXXV.As I noted in my Il libro Lambda, cit., p. 226 n. 9 (with cross references to ibid., § 2.2.2).Rashed 's  criteria  2,  6,  and  7  seem to be especially appropriate to the text of Metaphysics as well. 43Of course, things are now made more complicated by Oliver Primavesi's suggestion to explain some differences in textual extension between the two families as a presence of supplementary wordings in the one, rather than as a corresponding lacuna in the other.
[The top of the picture shows one of J 2 's marginalia: the words h] nou ' " kai; o[ rexi" kai; sw' ma at 1071a3, omitted by homoeoteleuton in J, are then supplied in margin by J 2 .Into evidence: the ligature εξ in J 2 .The bottom of the picture puts into evidence the different ligature the J scribe has for εξ, e.g. at 1071a9, to; ej x aj mfoi' n] Fig. 2 1069b34s.τὸ δὲ τρίτον ἡ ὕλη habuit J erasit et in mg.fecit J 2 | post ὕλη (a34) μετ ὰ-ὕλη om.J supplevit J 2 (Vind.gr. 100,f. 185 v ) [Here a transcription of the alleged two lines, L 1069b33-35.I underline the words omitted, again by homoeoteleuton, in J, then supplied by J 2 .They are in evidence in the picture: caiv , duv o me; n hJ ej nantiv wsis, h| " to; me; n lov go" kai; ei\ do" to; de; stev rhsis, to; de; triv ton hJ u{ lh.Meta; tau' ta o{ ti ouj giv gnetai ou[ te hJ u{ lh ou[ te to; ei\ do", lev gw de; ta; e[ scata.pa' n ga; r metabav llei The point at issue is also mentioned in the critical apparatus of my edition of Il libro Lambda, p. 173, ad ch.2, 1069b34, 35, and in my BMCR 2013.08.17 contribution, where I use as example the folios of Lambda, 185r-189v.There we see that J 2 has a script smaller in size, and is full of abbreviations, whereas J has none, but these difference could be seen as due to the fact that J2 is annotating as opposed to writing the main text (although it is hard to figure out while crucial errors like missing parts of the text should be supplied in smaller handwriting): however, more significantly, J 2 's letters are less round and regular; J 2 's ductus is more inclined; ligatures differ in the two hands (e.g.εσ, εξ); the letter ξ stands on the line in J, but goes down under the line in J 2 ; and so on.These remarks join in value a series of differences I pointed out in my critical edition, which had not been noticed by former editors, which I first detected by inspecting J in Vienna: by rasura, J 2 often corrects J's obvious errors.See most remarkably at 1069b34-5, where he supplies J's homeoteleuton (as a look at my app.cr., J 2 's intervention escaped the editor's attention at 1069a36 45 Special thanks are due to Carlo Maria Mazzucchi and to Enrico Berti for long lasting discussion on the whole subject matter; to Paolo Barattini and Helmut Woda for sensible thoughts and suggestions on this paper's structure; to Evan Keeling and Marco Zingano for their patience and accuracy in accepting for revision a particularly rough and late draft. Bernardinello's stemma, based on thousands of published collations from selected sections of the Metaphysics, is reproduced here below (L'eliminatio codicum della Metafisica di Aristotele, Padova 1970, p. 225): In 1972/1979, Dieter Harlfinger, who had already produced a stemma of the pseudo-Aristotelian treatise De lineis insecabilibus, undertook, under the request of Paul Moraux, a thorough rethinking of Bernardinello's volume and stemma.Based at the Aristoteles Archiv, founded by the same Moraux in 1965, Harlfinger was able to add several substantial contributions to the textual history of the Metaphysics, most of which are formulated and spelled out in the form of explicit or implicit disagreement with Bernardinello.
Still, perhaps, from this very figure, one could have thought already in 1979 that the boundary between alpha and beta was eventually to collapse.A radical rethinking of the whole tradition cannot be dealt within the present contribution but must remain at least as an open possibility.After Harlfinger's 1979 contribution, our research field has been sleeping for a long while.It took a few decades before an editor actually dared to edit such books based on such a stemma.

2.
Two stemmatically-based editions of single books of the Metaphysics (2012) This is true especially since 2012, when both Oliver Primavesi and Silvia Fazzo published editions of single books of the Metaphysics, Alpha 19 and Lambda respectively.As for Lambda, the project of an edition (2012) with commentary (2014) began in 2004 as a doctoral thesis, discussed in 2009. 20Meanwhile, a new edition of book Alpha was undertaken by Oliver Primavesi and his team on the behalf of the Symposium Aristotelicum committee.Fresh manuscript collations, based at the Aristoteles-Archiv (Freie Universität Berlin, dir.D. Harlfinger) were funded for this purpose.
establishment but has some weight for the history of the Metaphysics printed edition.What's the edition princeps manuscript Vorlage?What's the source of the Metaphysics 1497 editio princeps?Where textual history of the Metaphysics in early printed editions is concerned, the issue is second to none, because Aldus' text was largely reproduced in Erasmus' two-volume edition of Aristotle's works in all of its reprints.In the main substance, therefore, Aldus' Vorlage the basis of the reference text of the Metaphysics until Brandis 1823 and Bekker 1831.By the way, both Bernardinello and Harlfinger have the editio princeps of the Metaphysics as published in 1498.However, it was printed on 1 June 1497. 39This agreement in error a part, Harlfinger's view on the Aldina's Vorlage is different from Bernardinello's.