Democritus, Heraclitus, and the Dead Souls: Reconstructing Columns I–VI of the Derveni Papyrus [*]

Franco Ferrari
Università dell’Aquila
On January 15, 1962, the remains of the Derveni Papyrus were unearthed from a cist grave in northern Greece. Anton Fackelmann, curator of the papyrus collection of the Österreichische Nationalbibliothek in Vienna and the world’s leading expert in the handling of carbonized papyri, began work on the find. During the summer of that year, Fackelmann, working alone, succeeded in unrolling the fragments of the papyrus.
First he cut the scroll apart lengthwise, producing two symmetrical hemicylinders. Then he detached each of the layers of the two halves, placing them under seven pieces of glass (A–G). Beneath two additional pieces of glass (H and I), he arranged many fragments that he described in his report to the museum of Thessaloniki as “countless small pieces (down to a postage stamp) lying all around haphazardly” (see KPT, 5). These fragments consisted of about eighty bits of papyrus that had been labeled as “unplaced” on plates 28–30 of the editio princeps, in addition to the pieces placed by the first editors throughout the first twenty-six columns.
The outer parts of the roll—those that match the present columns I–VI, referring to prophecy and rites (prayers, offerings and libations)—had been reassembled on the basis of fragments from hemicylinders F and G, with occasional additions of pieces from the H and I groups. In turn, hemicylinder G consisted of two minor sections: “great G” and “small G” (Gg and Gp, respectively, in Figure 1).

Figure 1. Lettered volutes of the Derveni Papyrus.
In this configuration, each alternation of an F and a G section recomposes one circumference, or volute, of the roll, decreasing in size as we move from the outer (with a width of 9.8 cm) to the inner sections of the scroll and down to its midollo, or marrow. It is specifically on this basis that Tsantsanoglou (1997) reassembled these columns for the first time. His arrangement has been repeated in the editio princeps (KPT) with the exception of a change concerning G 7, which we will examine presently.
Therefore, one cannot say, as Janko (2008) has claimed, that Tsantsanoglou and Parássoglou ignored the procedures applied to the unrolling of burnt scrolls, wherein, as with the Herculaneum papyri, the bark must be cut into halves before the layers can be separated. Janko rightly noted, however, that the Greek editors made an unfortunate error in their edition, putting the piece G 7 in column II between two other fragments of the G type. In this way they mistakenly placed three sections of the G type in direct sequence.
This mistake did not derive from a disregard of the “alternation principle” regarding the F and G sections. It arose from the suspicion, clearly stated on page 64, that Fackelmann had wrongly classed G 7 as a fragment of the G, rather than the F, type. In fact, G 7 can be perfectly inscribed within the bottle-like surface of a “Great G,” which we see in its entirety (that of G 1) in a photograph taken before the separation of the G layers. This image, reproduced in the first plate of the editio princeps, shows the intact stalk with no overflowing under G 1. This means that all the pieces of the “Great G” type were inscribed within the perimeter of G 1. [1]
Moreover, the fact that G 7 belongs to the “Great G” series is corroborated by its formal congruence with a piece from the “small G” series, namely G 20, in that G 20 and G 7 create a “small G/great G” set (gG): hence, the possibility of reading, as I did in Ferrari 2011a:42–43, ἐλ̣[π]ίδι “to the hope” at line 4, ὧδ̣’ ἐ̣π̣έθηκε “he added in this way” at line 5, and μ̣[υ]σ̣τῶν “of the mystae” at line 7 (Figure 2). The actual fiber congruence between G 7 and the G 17/G 8 complex at the end of the first column can be explained at least as well as in the editio princeps if, with Janko (2008), we place G 7 at the beginning of the first rather than the second column (but at the height established by the first editors).

Figure 2. The reunification of G 20 and G 7 as parts of a “small G/great G” whole.
Another criticism of the Greek editors by Janko (2008) is that they ignored the kollesis that he claims is visible along the right edge of G 6. Solely on the basis of this alleged adhesion of two papyrus sheets, Janko disrupted the editio princeps’ arrangement of columns I–III, shifting G 11 from column III to column II and G 15/G 6 from column II to column III. Yet the kollesis Janko identifies would require, between this join within Janko’s column III and the undeniable kollesis recognized by the first editors at the end of column IV, a papyrus sheet (kollema) with the odd width of 12.85 cm. The widths of the other sheets in the first ten columns range from 14.6 to 16.25 cm. Also, in the photograph of G 6 taken through a microscope, a sottoposto (a scrap coming from a preceding volute and stuck under the higher level of the following one), rather than a kollesis, is visible along the right edge of the sheet. This observation is supported by the fact [2] that, in G 6, the upper layer reveals some extensions crossing the border of the edge (Figure 3). Furthermore, one can observe that the contour (G 6a) of G 6, which presumably (and, along its right edge, demonstrably) repeated that of its sottoposto, fits between G 17 and G 8 (Figure 4).

Figure 3. G 6 upper layer, revealing some extensions along the right edge.
Figure 4. The shape of G 6 fits between G 17 and G 8 (col. I).
Another change to the arrangement established by the first editors concerns the upper margin. In a comparison of plates 5 and 6 of the editio princeps, it is clear that the first line of G 1 in KPT is in fact the second line of the column (additionally, some ink traces can perhaps be identified in the photograph of the untouched stalk in KPT, plate 1). Therefore, if for practical reasons one would prefer to keep KPT’s numbering, one should add a line “0” at the beginning of each column.
In short, except for the unavoidable transfer of G 7 from the second to the first column and the insertion of a new line on the beginning of every column, the arrangement laid out by the first editors is trustworthy. As I hope to have proved in a new edition of columns I–VI, it allows a plausible re-creation of the argument developed by the anonymous Derveni author (henceforth D).
On the other hand, as I have already mentioned, some fragments belonging to the H and I groups might be assigned to the first columns. According to the first editors, these pieces include H 7 in column II, H 46 and H 8 in column IV, H 2 in column V, and H 18 and H 28 in column VI. Janko (2008) also adds I 70 in column VI along the left edge of G  2 (1.10).
Here, however, we should read not τὸν μέλλοντ]α θεοῖς θύειν with KPT, but ὁ μέλλων] ἱ̣ρὰ θεοῖς θύειν “he who is about to make offerings to the gods.” Janko writes ἱ]ε̣ρα, yet iota is almost certainly the first letter of I 17.2 (KPT, 124: “ι is very likely”), whereas epsilon cannot be read. Thus, the Ionic form ἱ̣ρὰ (cf. Herodotus 2.47.3 τὰ ἱρὰ θύσωσι) must be read (Figure 5) in spite of ἱεροῖ[ς] in line 6, ἱερά in column XX 4, and ἱερ[ο- in column VII 7. This is another instance of the continuous alternation of Ionic and Attic dialect forms in the text of the Derveni Papyrus. [3]

Figure 5. I 70 as placed by Janko along the left edge of G 2.
In his review of the editio princeps, Janko wrote that “it is astonishing to learn that 113 fragments, or 42% of the total, remain unplaced.” [4] Three days later the Greek editors replied in the same electronic journal that such a reproach reminded them “of the farmer whose only possessions were a bull and a frog and who, upon the demise of the latter, exclaimed in agony that he had lost half of his livestock.” Indeed, most of the unplaced pieces are very small. Statistically, however, it is very likely that quite a lot of them originally belonged to the first columns. Such probabilities diminish as one moves to the better-preserved columns.

Democritus and the First Lines of Column IV

First I would like to complete what I wrote in Ferrari 2010, where I suggested this reconstruction of column IV 0–4, starting with the fact that in column IV 4 (= fr. G 4.1) the sequence of letters ΟΥΤ in ]μμάνει̣ν̣; ἆρ’ οὐ τ.[ is shared on the upper-line level by the traces of the first three letters of an unplaced piece, F 14.4. In addition, another unplaced fragment, F 17, matches very well with H 46 and F 15, as well as sharing the epsilon of ἐὼν in 1.1 (Figure 6):
[ . ]ου ε . [. . . . . . . . . ὥ]σ̣περ φυσι̣κ̣[ὸς μετ]ὰ̣ δίκης ἐ̣ὼν
ὁ κείμ[ενα] μετα̣θ̣[εὶς] μὲν ἃ εὐχα̣[ῖς χρὴ] ἐκ̣δ̣ο̣ῦναι,
μᾶλλ[ον ἃ] σ̣ίνεται [ἢ ὡ]σ̣ ἀνημμέ[να εἰς] τ̣ὰ τῆς τύχης π̣[ῶς
4οὐκ εἴ̣ [α λα]μμάνει̣ν̣; ἆρ’ οὐ τα̣ῦ̣[τα κρατεῖ ο]ὐ̣δὲ κόσμος;
… he supposed … rightly being, like a physicist, … why, after altering the rudiments that should be given to prayers, did he not allow to consider what damages us more than something depending on chance? Isn’t it true that not even the universe is able to control it?
Figure 6. F 14 and F 17 placed at the top of col. IV.
The intellectual D charged with this controversy must have been Democritus of Abdera, born around 460 BC. In D’s opinion, Democritus had altered the basic principles of the traditional faith and cult. He had stated which of those principles should govern prayers and had refused to accept any unfortunate aspect of human existence as anything more than a chance accident. Actually, Democritus considered eidola and phantoms, perceived by common people as godlike phenomena, to be mental images (cf. Lucretius 5.1173–1174), atom clusters coming from the mind of man or beast and able to move through the air. He did not, however, underrate their impact. As we read in 68 B 166 DK (= Sextus Empiricus Adversus mathematicos 9.19), he prayed that he would encounter only benevolent eidola in his daily life (εὔχετο εὐλόγχων τυχεῖν εἰδώλων).
According to Democritus, the wise man (the ἀνὴρ εὐλόγιστος of 68 B 236) can hope that the eidola disseminated through the air do not interfere with his peace of mind (cf. Epicurus Letter to Herodotus 49). If they do, however, he cannot recover his mental or bodily health by entreating the gods (68 B 234). This sole Democritean argument allows D to state that Democritus altered (μεταθείς) religious habits and beliefs by arguing that the harmful aspects of life did not depend on the gods, but on atomic determinism.
My reconstruction entails two facts that seem to be at odds with the general principles established by the first editors:

  1. (1) The first line of the column is actually immediately above the line identified by KPT as the first.
  2. (2) The fragment F 14 in fact belongs within a G section.

The first point presents no problem since, as we saw above, the editio princeps fails to recognize that the actual first line of every column is directly above the line it misidentifies as the first.

On the second point, until recently I believed I had overcome the obstacle by acknowledging that, in accordance with Fackelmann’s report and Tsantsanoglou’s clarifications (KPT, 4–5), the G series includes two different typologies. First are the fragments which allowed Fackelmann to reassemble the outer volutes of the roll: the pieces F 1–9, D 1–11, and E 1–13 (Fackelmann’s “third type”) alternating with the pieces G 1–21 (Fackelmann’s “fourth type”). Second are the fragments F 10–20, H 1–68, and I 1–97, which comprise the “countless small pieces lying all around haphazardly” that we recalled above.
A new communication to M. S. Funghi from the first editors stating that the separation of these two typologies arises only from an oversight of theirs would seem to run against my theory. However, in reexamining F 14, I noticed that this piece is actually made of two smaller bits, rightly put together by Fackelmann before being placed under glass. The left fragment (Fa) fits together very well with G 4, a “great G” piece (Figure 7). Thus, Fackelmann’s assignment of F 14 to the F group seems to have been a mistake. [5]

Figure 7. Suggested placement of F 14a within a “Great G” piece such as G 1.

Heraclitus, the Persians, and Column IV 7–14

In column IV, lines 5–9, D names Heraclitus as one who, like Orpheus, spoke as a hierologos, giving credence to broadly accepted feelings or opinions (κοινά) at the expense of idiosyncratic ones (ἴδια). The quotation (perhaps a little free, as I have shown in Ferrari 2011b) encompasses what were thought to be two separate Heraclitean fragments (22 B 3 and B 94 DK), as two different authors quote them. Thus we can suppose that these fragments initially belonged to a single statement.
In spite of the papyrus gaps, the Heraclitean quote has been adequately recovered through other evidence. The differences between the first editors on one side and Janko (2002) and Bernabé (2007) on the other concern the formal shape of the passage rather than the substance of the argument (col. IV 5–11, with Sider’s ἱερο]λόγωι in line 6, my supplement περιό]δ̣ου in line 7, and Lebedev’s θύο̣υ̣[σι in line 11; the other supplements are by the first editors).

5κατὰ̣ [….]α̣ Ἡρ̣άκλ̣ε̣ιτοσ μα̣[. . . . . . . . . ] τ̣ὰ κοινά,
κατ̣[αστρέ]φ̣ει τὰ ἴδ[ι]α, ὅσπερ ἴκελ̣α̣ [τῶι ἱερο]λόγωι λέγων [
“ἥλι̣[ος περιό]δ̣ου κατὰ φ̣ύσιν ἀν̣θρω[πηΐου] εὖρος ποδός [ἐστι
τὸ μ̣[έγεθου]ς̣ οὐχ ὑπερβάλλων εἰκ̣[ότας οὔ]ρους ε[ὔρους
ἑοῦ· εἰ δὲ μ]ή, ̓Ερινύε̣[ς] νιν ἐξευρήσου̣[σι, Δίκης ἐπίκουροι
10ὑπερ]βατὸμ ποῆι κ[
]α̣ι̣ θύο̣υ̣[σι
… invoking common truths Heraclitus disrupts the idiosyncratic opinions, he who said, speaking like an author of sacred tales: “The sun, according to the nature of its circumference, is a human foot in width, not exceeding in size the proper limits of its width. Or else the Erinyes, assistants of Dike, will find it out … so that it makes not passable … they make sacrifices …
One gathers from the references to a human foot, its “size” (εὖρος) and its “boundaries” (οὔ]ρους) as terms of comparison, that the issue here is the real (κατὰ φύσιν), rather than the apparent, size of the sun. [6] In the next line ὑπερ]βατὸμ ποῆι must be part of a statement by D, rather than Heraclitus, where the subjunctive ποῆι is the core of a final clause connected with a sentence pivoting on θύο̣υ̣[σι.
D himself (rather than Heraclitus) reminds his audience that some people (their identity is lost in the gaps of the papyrus) offer sacrifices to the sun in order to prevent it from defying its own physis by changing size. The identity of this people can be recovered through a neglected palaeographic datum: two tiny fragments, so far left unplaced (I 62 and I 80), can be put together using both their shape and the trend of their fibers. The resulting larger piece fits easily between G 4 and H 8 (Figure 8).

Figure 8. The new placement of I 62, I 80, and H 66 in relation to G 4 + H 8 (col. IV).
Hence, with Piano’s ingenious supplement to line 10, the new text of lines 10–13 is:

10μὴ ἑὸμ μέγεθος ὑπερ]βατὸμ ποῆι κ[
Π]έ̣ρ̣σ̣α̣ι̣ θύο̣υ̣[σι(ν)
κ]ατ̣ὰ̣ τ̣ὰ Δίκης [μέτρα (?)
γὰ]ρ̣ [ἀ]μήνιτα κ[
… Persian people make sacrifices … so that (the sun) does not make surmountable its own size … according to Dike’s rules … not angry indeed …

Thus, taking into account the interest in Iranian ritual practices and the entities to whom they were addressed, the sacrificers must be Persian people. [7]

Finally, another quote from Heraclitus (22 B 52):

αἰ]ώ̣ν̣ ἐ̣σ̣τ̣ι̣ π̣α̣ῖς̣ π̣[αίζων, πεσσεύων· παιδὸς ἡ βασιληίη
Eternity is a child playing at the table: the kingship of a child.

comes to light in line 14 thanks to one more tiny unplaced piece, H 66, that can in fact be placed below I 80 and H 8. [8] The outline of the higher right edge of H 66 matches with that of the lower left edge of H 8. Also, the two pieces share the sigma and tau of ἐστί. As to the deciphering of the letters of H 66, ε̣ς̣ has been already recognized by the first editors (KPT, 120) and can be taken as certain, but the sketch of the preceding ΩΝ also seems retraceable (KPT, pl. VIII, detail).

The meaning of this puzzling passage must be, at least according to D, that αἰών, the ceaseless stream of time in the universe, is like a child playing at petteia (a board game of controversial identity). Because the scene is repeated in every daily or annual cycle, it is, paradoxically, forever both old and young. “The everlasting child,” as Kahn wrote, [9] “remains forever youthful, even infantile … playing his endless game and maintaining kingship by a series of births and deaths across the generations” (even the gods play at petteia with men’s souls as their pieces in Plato Laws 903d.)

The Souls of the Right Ones and the Beginning of Column VI

My last issue concerns the beginning of column VI, which deals with rituals (spells, offers, libations) undertaken both by magi and by mystae.
Here is the text of lines 1–3 as it was established by the first editors:

[ ca. 8 εὐ]χ̣α̣ὶ καὶ θυσ̣[ί]α̣ι μ[ειλ]ίσ̣σ̣ο̣υσι τ̣ὰ̣[ς ψυχὰς,
ἐπ̣[ωιδὴ δ]ὲ̣ μάγων δύν[α]ται δ̣αίμ̣ο̣νας ἐμ[πο]δ̣ὼ̣ν̣
γι̣[νομ]έ̣ν̣ο̣υ̣ς μεθιστάν̣αι…
… prayers and sacrifices appease the souls, while the incantation of the magi is able to drive away the daimones who are hindering …
(Translation in KPT)
The initial point to be noted here is that the first, partly legible word is not εὐ]χ̣α̣ί but, as Tsantsanoglou (2008) has recognized, χ]ο̣ή̣ “libation.” This reading establishes a connection between χοή and θυσίαι, one closely paralleled on the Persian side in Xenophon’s Cyropaedia, where Cyrus the Great, just after crossing the Assyrian frontier, offers libations to appease the land and its gods and “heroes” (3.3.22). Herodotus (7.43) had already mentioned libations offered by Xerxes to the “heroes” of the Troad. [10]
A more subtle issue is the supplement to be adopted at the end of line 1. Tsantsanoglou [11] rightly remarked that Persian sacrifices in the presence of a magos were addressed to ancestors of a family group (cf. Herodotus 7.191; Strabo 15.3.14; Pausanias 5.27.5). He also noted that the Iranian equivalents of Greek ancestors or “heroes” were the artavan or ashavan, “those who possess truth.” These ancestors are mentioned in some Greek sources as artaioi or artades: Herodotus 7.61; Hesychius α 7472 Schmidt ἀρτάδες· οἱ δίκαιοι, ὑπὸ μάγων and 7473 ἀρταῖοι· οἱ ἥρωες, παρὰ Πέρσαις; Stephanus Byzantinus Ethnicorum quae supersunt, p. 127 = Hellanicus FGrHist 4 F 60. The feminine ἀρτάδες can depend on the fact that by this word the source of Hesychius probably pointed to the fravashi (this Avestic word is grammatically feminine) of the artaioi, namely, their souls.
On the other hand, the daimones of line 2, beings to be driven back rather than appeased, are opposed to the artades. They should be identified not with the fravashi, as Tsantsanoglou presumed, but with the daêvas (recorded in Hesychius’ entry δ 714 Schmidt Δεύας· τοὺς κακοὺς θεούς. Πέρσαι), who were perpetually at war with the fravashi. The daêvas were to be driven off from the places where they lurked, trying to hinder the deceased on their way toward the afterlife. [12] Against them the magi sang their epoidai, [13] Avestic texts that were incomprehensible to Greek ears, sung in low voices.
Thus, if the daimones can be identified with the daêvas of Persian religion, a supplement such as τ̣ὰ̣[ς ψυχὰς (KPT) is not a suitable name for these beings, who should be designated in a more specific way than with the generic word “souls.” But, precisely that, the Iranian loan ἀρτάδες with the explanation in the Hesychius’ gloss οἱ δίκαιοι “(the souls of) the Right ones,” occurred at the end of line 1 if one places the piece I 24 in the upper right corner of this column (Figure 9). The fiber consistency between G 2 and I 24 is striking, and the reading τ̣ὰ̣[ς ἀ]ρ̣τά̣δ̣α̣[ς is corroborated by the obvious supplement ἐμ[πο]δ̣ώ̣ν in the following line (already suggested by KPT on the basis of ἐμπο[δών in line 3).

Figure 9. I 24 placed near G 14 + G 2 (col. VI).
Therefore, we may have identified the correct word for the recipients of the libations and sacrifices of the Persian people.


Bernabé, A., ed. 2007. Musaeus · Linus · Epimenides · Papyrus Derveni · Indices. Poetae epici Graeci: Testimonia et fragmenta, part II, fasc. 3. Berlin.
Carastro, M. 2006. La cité des mages: Penser la magie en Grèce ancienne. Grenoble.
de Jongh, A. 1997. Traditions of the Magi: Zoroastrianism in Greek and Latin Literature. Leiden.
Ferrari, F. 2010. “Democrito a Derveni? PDerv. col. 4, 1–6.” Parola del Passato 65:137–155.
———. 2011a. “Frustoli erranti: Per una ricostruzione delle colonne 1–3 del Papiro di Derveni.” In Funghi 2011:39–54.
———. 2011b. “Rites without Frontiers: Magi and Mystae in the Derveni Papyrus.” Zeitschrift für Papyrologie und Epigraphik 179:71–83.
Funghi, M. S., ed. 2011. Papiri filosofici: Miscellanea di Studi VI. Florence.
Horky, P. S. 2009. “Persian Cosmos and Greek Philosophy: Plato’s Associates and the Zoroastrian Magoi.” Oxford Studies in Ancient Philosophy 37:47–103.
Janko, R. 2002.The Derveni Papyrus: An Interim Text.” Zeitschrift für Papyrologie und Epigraphik 141:1–62.
———. 2008. “Reconstructing (Again) the Opening of the Derveni Papyrus.” Zeitschrift für Papyrologie und Epigraphik 166:37–51.
Kahn, C. H. 1979. The Art and Thought of Heraclitus. Cambridge.
KPT = Th. Kouremenos, G. M. Parássoglou, and K. Tsantsanoglou, eds., The Derveni Papyrus, Edited with Introduction and Commentary (Florence, 2006).
Lulli, L. 2011. “La lingua del papiro di Derveni: Interrogrativi ancora aperti.” In Funghi 2011:91–104.
Marcovich, M., ed. 1978. Eraclito. Frammenti. Florence.
Panaino, A. 2009. “Aspetti della complessità degli influssi interculturali fra Grecia e Iran.” In Grecia Maggiore: Intrecci culturali con l’Asia nel periodo arcaico. Atti del 75 anniversario di W. Burkert, Istituto Svizzero di Roma (2 febbraio 2006), ed. Ch. Riedweg, 19–53. Basel.
Piano, V. 2011. “Per una revisione papirologica delle prime colonne del papiro di Derveni.” In Funghi 2011:5–37.
Tsantsanoglou, K. 1997. “The First Columns of the Derveni Papyrus and Their Religious Significance.” In Studies on the Derveni Papyrus, ed. A. Laks and G.  W. Most, 93–128. Oxford.
———. 2008. “Magi in Athens in the Fifth Century BC.” In Ancient Greece and Ancient Iran: Cross-Cultural Encounters, ed. S. M. R. Darbandi and A. M. L. Zournatzi, 31–39. Athens.


[ back ] * The author’s own edition of the papyrus may be found at
[ back ] 1. See Piano 2011:18–19.
[ back ] 2. Highlighted by G. Del Mastro ap. Piano 2011:34–35.
[ back ] 3. See Tsantsanoglou in KPT, 11–14 and Lulli 2011.
[ back ] 4. Bryn Mawr Classical Review, 2006.10.29.
[ back ] 5. For another instance (the erroneous classification of D  12–14), see Piano 2011:10.
[ back ] 6. See Kouremenos in KPT, 156–161.
[ back ] 7. As testified during the Classical age by the sources (Herodotus 1.131.2; Xenophon Cyropaedia 8.7.2). A plain reference to them also comes out in Plato Laws 821c–d, where the Athenian urges the reader to honor “the gods of the sky” (see Horky 2009:48).
[ back ] 8. The word order παῖς παίζων occurs in Lucian Vitarum auctio 14; see Marcovich 1978:339.
[ back ] 9. Kahn 1979:228.
[ back ] 10. See Carastro 2006:17.
[ back ] 11. Tsantsanoglou 1997:111–112 (and 2008:33).
[ back ] 12. See Panaino 2009:34–36.
[ back ] 13. See Herodotus 1.132 and, above all, for spells performed against numinous beings, 7.191: κατα-είδοντες βοῆισι.