An early version of this article was presented at the 30th International Congress of Papyrology, prior to the publication of Perrone 2022. It has been revised with an eye to minimizing the reduplication of information from that study.
My work has benefited from others’ generosity and support, and I thank especially Mark Depauw and Yanne Broux (Trismegistos / Leuven), Monica Tsuneishi (Michigan Papyrology), Michelle Fontenot and Tracey Miller (Kelsey Museum of Archaeology), Anne Harlow and Moira White (Tūhura Otago Museum), Evelyn Adkins (Case Western Reserve University), and Christine Colburn (University of Chicago Library), as well as my collaborators Nicholas Hudson (UNC Wilmington) and Liwei Zheng (University of Manitoba). The letters of Orlando W. (Pip) Qualley are quoted by permission of the Luther College Archives, Decorah IA, where they are kept (RG15, Box 8). Financial support for the research was provided by the Social Sciences and Humanities Research Council of Canada as part of the research project ‘the Books of Karanis’.
In a recent article, Serena Perrone has mined the Homeric papyri from Karanis for evidence about Greek literary culture there, employing an interdisciplinary approach whose contextualization of the texts incorporates legacy data from the University of Michigan’s 1924–1935 excavations at the village. Using the dig’s invaluable record of excavated objects, Perrone surveys the distribution of Homeric papyri by their stratigraphic level to reassess what we can learn about readership in an ancient village. She notes repeatedly the archaeological caveats associated with the excavation’s methodology as well as how tenuous the attempts to connect individual fragmentary texts with specific readers or libraries have proven. E.g., Perrone 2022: 37, 41–42, 45.
This article supplements Perrone’s findings by bringing additional archival and archaeological data to bear on the contextualization of the Homer papyri from Karanis – both purchased and excavated. The latter work proceeds in parallel with traditional papyrological and palaeographical (re)analyses of the texts, but it is nonetheless conducted independently of them, in the interest of methodological control. As Perrone rightly notes, the levels recorded by the Michigan excavators are not stratigraphic levels in the real archaeological sense,
Perrone 2022: 41. E.g., Perrone 2022: 55–56.
Of the ca. 2,000 Homeric papyri in the Leuven Database of Ancient Books (LDAB), thirty have ‘Karanis’ as their stated provenance. For the first figure, cf. Perrone 2022: 37, n. 1. The latter was retrieved via a search in LDAB for ‘Homer’ in the field ‘Ancient Author & Work’ and ‘Karanis’ in the field ‘Provenance’, last accessed on 1 July 2025. The LDAB is now a component of Trismegistos: Depauw and Gheldof 2014. P.Mich. inv. 5451 (TM 59200). P.Mich. inv. 4832 (TM 61219) So also Perrone 2022: 46 n. 51. The papyrus is P.Mich. inv. 15 + 15bis (TM 60685), which belongs to the Michigan collection’s founding purchase of 534 papyri: see Boak 1922: 38, Shelmerdine 1975: 21–22, no. 3, and Schwendner 1988: 13, no. 5, none of which mention Karanis. The acquisition report prepared for the collection by Arthur S. Hunt is similarly silent as regards provenance. On Kelsey and Grenfell’s collaboration, see Todd Hickey’s forthcoming introductions to P.Mich.Cent. and P.Lond. 8.
Brasch spent three seasons in Egypt in 1932–1935, working ‘as a kind of unpaid cadet’ on the Egypt Exploration Society excavations at Tell el Amarna, at the same time as C. H. Roberts, his Oxford contemporary, was engaged as a papyrologist with the University of Michigan expedition at Karanis. There is no information on the provenance of the texts published here, but Brasch visited Roberts at Karanis several times, and the pieces may have come from there since the lessee in (3), a landlease, comes from Arsinoe.
Whitehorne 1983: 39.
The information in the anecdote is an insufficient basis for even a tentative assignment of provenance. If one excludes these as well as the three items of Homerica, a census of the Homer papyri from Karanis more accurately begins with the following twenty-five individual copies of the Cf. Perrone 2022: 58–61, who includes among the list of Homerica on p. 60 two additional items.
The remarks in the preceding table under ‘Acquisition’ are sufficient to demonstrate the various books’ origins in Karanis. The final twenty items (Nos. 6–25) were all excavated by the University of Michigan (1924–1935), and their provenience is accordingly documented with a field number. Those Michigan papyri fall into two groups. Eleven of them (Nos. 6–16) were uncovered during the third season of work (i.e., 1926/7) in the neighborhood dubbed ‘Area G’ by the excavators. On the Homer fragments from Area G, see also now Sampson 2023: 462–464.
Like the first three items in Table 1 (which are treated in more detail elsewhere See Sampson, Sawatzky, and Shields 2025, in this volume. “(1) The papyri were found in the Faiyum at a town called Washim. (2) They were found by people digging in the heaps or ‘tells’ of dirt and rubbish at the edge of the town. As they dug down they found walls of houses standing in position. In the rooms of these houses or some of them they found the papyri – some were in sacks, others in piles in corners, etc. (3) The pieces are all from that one town but probably not all from the same old ruin – but from different houses.” J. R. Alexander to E. J. Goodspeed, 14 March 1898, reproduced by Hickey and Keenan 2021: 17, quoting a February 1973 transcription by R. W. Allison now in the Goodspeed Collection at Denison University.
Goodspeed posited a date of “not later than 159” in his
Goodspeed 1898: 348.
Goodspeed 1898: 349. The Goodspeed papyrus includes scribal accents (acute, grave, circumflex), apostrophes, and ano-stigmai.
Comparison to that famous example of a formal round is generous. Although phi often violates bilinearity (while rho and iota do not), the script of Goodspeed’s papyrus is squatter and its variation between broad (alpha, lambda, nu, omega, tau, pi), square (omicron, theta, eta, sigma), and narrow letters (gamma, rho, iota) is far more pronounced. The lower extremities of many letters end in a serif-like blot, as though the scribe let the pen rest momentarily. In its general appearance, the capitals aspire to the rounded majuscule type, but their execution is quite stiff and imperfect. It is an early hand, and a date in the first century is preferable; cf. P.Oxy. 2 246 (66); PSI 1 11 (middle I). For the former, see Cavallo and Maehler 2008: no. 96; for the latter see Cavallo and Maehler 2008: no. 95.
Cavallo 1975: 32.
No image of P.Chic. 5 was published, and its present whereabouts are unknown. See Allison 1975: 32. See Goodspeed 1906: 167; the provenance is reiterated in P.Chic. 5. Perrone 2022: 45, n. 44 is more cautious.
During the first two excavation seasons, the Michigan team had focused on several regions of Kom Aushim that it identified by the letters A through F. See Boak and Peterson 1931: 3–4, and (for an illustration) Plan 1. The precise location of Area G remains unknown: for a description, see Boak and Peterson 1931: 7–8. Cornelia Römer has argued that, in light of the size of the structures and the tendency of the Fayum wind to blow from the northwest, Area G was a prosperous neighborhood (P.Cair.Mich. 3, p. x–xi). On the background to the third season, see Sampson 2023: 457–459.
The eleven Homeric papyri discussed below are a small fraction of the bounty of texts uncovered in Area G. These included, most notably, portions of the so-called ‘archive’ of Sokrates the tax-collector, which was treated by Peter van Minnen in a study that has proven foundational to papyrology’s interdisciplinary turn.
van Minnen 1994, building upon Strassi Zaccaria 1991: 245–262. See also Strassi 2001 : 1215–1228.
Sampson 2023: 453–464. So also Perrone 2022: 54.
Most’s On No. 1, see Sampson, Sawatzky, and Shields 2025: §4. So also Perrone 2022: 53. Other unpublished texts from House B7 are P.Mich. inv. 4765 (26-B7A-A); P.Mich. inv. 4769 (26-B7L-B); P.Mich. inv. 4770 (26-B7N-CI); and a third-century ostracon (26-B7N-E). Another third-century ostracon from the house (26-B7N-F) was published as O.Mich. 1 147.
The field number assigned to the papyrus by the dig’s record of objects (26-B7H-H) indicates the season of excavation (1926/7) as well as the layer, structure, and internal subdivision (House 7, from the ‘B’ layer, Room H). The final element in the field number itemizes the finds within a subdivision: H, in this case, indicates the eighth item recorded from Room H. House B7 is located on the eastern side of the southeastern quadrant of Area G (see Figure 4).
Although House B7 warrants mention in the excavation’s architectural report solely for the white plaster pointing of its bricks and its large, bricked-over oven, For the plaster pointing, see Boak and Peterson 1931: 29; for the oven, see Boak and Peterson 1931: 36–37, Fig. 54. Access to House B7 could also be gained from the north, as the caption to 5.2380 implies: “View looking southwest over portions of B3, 4 and entrance to B7.” See Figure 6. The designation of these two subdivisions as a corridor and staircase, respectively, does not occur in the Kelsey Museum’s photographic archive, but in the handwritten caption to photograph 5.2289 in an archival folder created by Sam Yeivin: “Impression of wooden threshold beam between corridor in h. B7A & stairway B7C.” See Folder 7, Box 17, Karanis Excavation Records, Kelsey Museum of Archaeology, University of Michigan. Caption to 5.2484: “Passageway, BS7, showing the wooden beam ends in the western wall of B7D at floor level between B7D and G.” We interpret the caption’s reference to the B7D’s ‘western’ wall as an error for ‘southern’, since Passageway BS7 (from which the photograph was taken) bordered House B7 to the south.
Analysis of the photographic archive also clarifies several aspects of Courtyard B7F, which Figure 4 indicates was situated between House B7’s main structure (to the south) and House B3 (to the north). Caption to 5.2016: “View to the north showing the walls of B1 and B2, street BS3 in the background and the walls of B7F in the foreground, before excavations. To the north of B7F lies B3 with BS1 between B3 and B1-B2, as they appeared before excavations.”
Figure 9 and Figure 10 present different perspectives on a niche in the northern wall of B7F, and both photographs’ captions indicate that the shot looks into B7N, the chamber on the left which is identifiable by the ovens sunk into its floor. These are the same ovens as those depicted in 5.2091, whose caption reads: “Ovens in B7N.”
Moving clockwise from the north wall of B7F, Figure 11 presents a view of Courtyard B7F’s eastern wall. The frame is centered on Alcove B7K, atop which Stairwell B7L ascended from the south. See Figure 11 and Figure 12. The caption to 5.2319 indicates that the eastern wall of B7E had a niche, but that description is incompatible with the visual testimony of Figure 11 and Figure 12, and another wall of the structure is presumably meant. Caption to 5.2070: “Fallen roofing in B7.”
Much becomes clear about House B7 from a closer analysis of the legacy date (see Figure 13). Nonetheless, despite their richness, the preceding analyses are insufficient to determine the precise location of Room B7H – the Homeric papyrus’ findspot. My hypothesis is that B7H was adjacent to (or a further subdivision of) Room B7G in the structure’s southwestern corner, but that hypothesis is grounded chiefly in the fact that texts were excavated in both B7G and B7H and that alphabetical order is compatible with (but not definitively constitutive of) proximity.
P.Cair.Mich. 2 1 and P.Cair.Mich. 2 2 were excavated from House B17 in the northwestern quadrant of Area G, and were part of a large cache of ca. 142 papyri that was assigned the field number 26-B17F-A in the dig’s record of objects.
Sampson 2023: 453 n. 62. On House B17, see especially van Minnen 1994: 237–244.
van Minnen 1994: 240; cf. van Minnen 2010: 463. So also Perrone 2022: 52. For texts from House B17 that do not obviously pertain to Sokrates or members of his family, see Sampson 2023: 460–461 nn. 89, 97.
Papyrological finds were not limited to Vault B17F: Room B17E also yielded papyri, though for these the excavation’s record of objects includes the annotation “papyrus very high in debris above the room,” evidently in acknowledgement of an archaeological abnormality. B17E had brick floors that were not covered in mud, but its location within the home is unclear: Room E had brick floors that were not covered in mud, into whose northwestern corner a pot was sunk (see Boak and Peterson 1931: 27, and Fig. 40; as well as the caption to 5.2137: “A view of the northwest corner of B17E showing a small pot sunk into the brick floor. This photograph was taken before the room was fully cleared.”). One presumes that it was located on the building’s ground floor and that it was adjacent to Room F, where the large cache of papyri which included the Homer fragments was discovered; there is no evidence of staircases or upper stories on Houses B1 or B2, which were located across Street BS2. See caption to 5.2137, quoted in fn. 40.
The script of P.Cair.Mich. 2 1 can be characterized loosely as an informal round. Its epsilon is narrow but for a high and wide middle bar. Kappa’s arms are also broad, as is the flat top of pi, which extends outside each hasta. Small serifs appear atop delta, eta, kappa, and sometimes iota and nu. It is not unlike P.Oxy. 60 4009 (II). The script of P.Cair.Mich. 2 2 is a somewhat more rapid example of an informal round; the rightmost extremities of letters frequently touch the left edge of the subsequent letter, though not in a way that could be called semi-cursive. Characteristic of the hand is epsilon, whose crossbar is broad and descends slightly to the right; one might compare P.Oxy. 31 2536 (II); PSI 15 1470 (II/III).
P.Cair.Mich. 2 6 was one of a cache of over thirty papyri from the second and third centuries which was assigned the field number 26-BS1-PI. These papyri were found not inside a structure but rather outside, on Street BS1, the east-west thoroughfare at the heart of Area G. An APIS search for the acquisition ‘BS1’ returns 56 hits, but several of these are duplicates that only distinguish the obverse sides of a single papyrus. For 26-BS1-PI, the dig’s record of objects includes the interesting annotation “Note papyrus found 1932–33”, which is explained by the appearance in APIS of the field number 32-BS1-PI for the fragments of P.Mich. inv. 6292: this field number, however, does not appear in the record of objects. Published papyri from Street BS1 are P.Mich. 6 366; P.Mich. 6 388; P.Mich. 8 495; P.Mich. 8 505; P.Mich. 8 512; P.Mich. 9 522; P.Cair.Mich. 2 12a; P.Cair.Mich. 2 12b; P.Cair.Mich. 2 12c; P.Cair.Mich. 2 12d; P.Cair.Mich. 2 12e; P.Cair.Mich. 2 16; P.Cair.Mich. 2 18; P.Cair.Mich. 2 19r; P.Cair.Mich. 2 19v; P.Cair.Mich. 2 20; SB 12 11104; SB 12 11124; SB 14 12082. These date to the second and third centuries. I.e. the many fragments of 2-A), but the field numbers beginning 26-BS1 are unusual inasmuch as they lack any further subdivision (as would be expected in a structure) – Street BS1 is a major thoroughfare in Area G! The field numbers for this context, rather, appear simply to itemize the finds, and every papyrus has the field number 26-BS1-PI. The hypothesis that a single cache was uncovered in a particular location on the street is further weakened by the fact that some literary fragments from this context join with others found in nearby structures, providing clear evidence that, as has been documented elsewhere, the papyri in Area G have ‘wandered’ and that the circumstances of their deposition are indeterminable.
The hand of this papyrus is a clear informal round: The
House B25 yielded fragments of three separate bookrolls of the
Boak and Peterson 1931: 7, and Fig. 3. See also my Figure 17 and Figure 18.
Boak and Peterson 1931: 34–35, Figg. 50, 51.
Römer’s See Turner 1987: no. 48.
The record of objects notes that P.Cair.Mich. 2 8 (26-B25T-C) was uncovered “high in filling,” an annotation which indicates the same discomfort with the archaeological context as for the papyri of Room B17E. The fact that B25 was covered by several meters of sebakh – especially, as Figure 17 indicates, along the western wall where this chamber was located – makes that annotation less surprising, and undermines the possibility of a clear relationship between text and context. One ought not to assert that the papyrus was found in the bin. Its script is an elegant formal round, the small sample size of which makes it somewhat difficult to determine whether it is an example of biblical or rounded majuscule. Despite the horizontal bar of alpha, its bilinear rho, and the presence of serifs on alpha at both apex and base of the right diagonal (all of which would suggest the latter), Römer’s In editing the Oxyrhynchus papyrus, Edgar Lobel described its hand as ‘biblical’, but it contains some elements that are foreign to that style: see Cavallo 1967: 31.
The hand of P.Cair.Mich. 3 3 (26-B25A-A), meanwhile, writes neat and round informal capitals, with serifs on the apices or bases of letters (especially of kappa). Characteristic letter shapes include the eyelet epsilon, h-shaped eta, and very wide nu and omega. The horizontal of tau often touches the preceding letter, and the right leg of pi is gently curved. I compare it with P.Oxy. 26 2441 (mid II) and would assign a date in the second century CE.
House B12 yielded two copies of Homer, which were both recorded under the field number 26-B12L-A. Figure 4 distinguishes a north complex and a south complex of House B12, separated by a courtyard, but makes very little clear about the structure’s layout beyond these basic subdivisions. The house lay on the northwest corner of the insula of buildings in the southeastern quadrant of Area G, bordered by Street BS1 (to the north) and Street BS2 (to the west). The excavation records indicate that B12L was the central courtyard, that it yielded more excavated objects than any other context in this structure, and that – at one time, at least – it was accessed from Street BS2. See Figure 22. That the doorway from BS2 into B12L was blocked is owed to the handwritten caption to photograph 5.2216 in an archival folder created by Sam Yeivin: “Entrance into house B12, courtyard, from BS. Note stone paved area in front of blocked doorway.” See Folder 7, Box 17, Karanis Excavation Records, Kelsey Museum of Archaeology, University of Michigan.
Archival photographs add to our understanding of the architecture: Figure 23 indicates that Room B12A was situated along BS1 and therefore belonged to the northern complex. See Figure 23. P.Mich. inv. 4784, an unpublished papyrus dating to the reign of Marcus Aurelius, was excavated in this room (26-B12A-AI). See Figure 24. P.Mich. inv. 4785 and 4826, unpublished fragments from the second and third centuries, respectively, were found in B12B (26-B12B-L). B12G, meanwhile, yielded seven coins, whose dates range from Ptolemy IV (222–203 BCE) to Trajan (109 CE) (26-B12G-H). See Figure 25. These details, which are sufficient to locate B12M in the southern complex of B12, are owed to the handwritten caption to 5.2082 in an archival folder created by Sam Yeivin: “House B12M, view looking SE, showing also the N. & W. walls of h. B55 in middle background. In far background – cultivation.” See Folder 7, Box 17, Karanis Excavation Records, Kelsey Museum of Archaeology, University of Michigan. See Figure 26. We suggest that the square entranceway at the top of the image is the same one visible in the bottom of Figure 25. Published texts with the field number 26-B12L-A are SB 24 15877, SB 24 15887 , SB 24 15864, SB 22 15864, SB 22 15789, SB 22 15786, ChLA 42 1214, P.Lond.Lit. 166 + P.Cair.Mich. 2 10, and P.Lond.Lit. 58 + P.Cair.Mich. 2 11. Two third-century ostraca were also found in Courtyard B12L: O.Mich. 1 331 (26-B12L-BI) and O.Mich. 1 375 (26-B12L-QII).
The first of the Homeric papyri from Courtyard B12L is P.Cair.Mich. 3 2, written in rounded capitals of an informal but decorated type. Lambda, hypsilon, and mu are upright, while nu and delta are wide, as is omega, whose middle element reclines slightly. The upper arm of kappa is both broad and nearly horizontal, thereby hazarding confusion with eta. The bubble of alpha is small and the tail gently curved, while delta has a pronounced cap and a curved left diagonal. Bilinearity is observed except for the tail of phi and a somewhat shallower right hasta of eta. Cf. BKT 9 10 (II); P.Oxy. 3 454 (later II). Courtyard B12L also yielded two fragments of P.Cair.Mich. 3 4. A third fragment of this papyrus has the field number 26-B2W-C: it was found in B2, a house on the north side of Street BS1, opposite House B12. For the layout of B2’s underground chambers, see Boak and Peterson 1931: Plan IIIA, and the general discussion of pp. 20–23. The wandering of P.Cair.Mich. 3 4’s fragments is consistent with the evidence that this part of Area G had been disturbed already by 1891. P.Cair.Mich. 2 10 + P.Lond.Lit. 166 and P.Cair.Mich. 2 11 + P.Lond.Lit. 58 originate from the same roll, a portion of which was excavated and assigned the field number 26-B12L-A, while the other was part of the lot acquired from the Rev. Greville John Chester on 2 February 1891. See Sampson 2023: 451–456.
House B9 lay to the immediate east of House B12. It was noteworthy for its dovecote B9P,
Boak and Peterson 1931: 24–25, Figg. 30–32. The photograph 5.2154 is overexposed, but its caption reads as follows: “Looking southeast over a portion of B9 in process of excavation. The blocked up doorway and the window in the western wall of B10 are just uncovered. B9N, M and W are just in process of being dug out.” See n. 58, above, for the caption to 5.2154.
Other aspects of the structure’s layout are harder to pin down. Because of its prominence, the dovecote is frequently a reference point for other subdivisions: it bordered on Room B9G, See Figure 29. See Figure 32. See Figure 33. The photograph 5.2482 is not especially clear, but its caption reads as follows: “Bolt case and bolt used for doorway in B9E, where B9E turns to the left.”
As was noted, two Homeric papyri were excavated in House B9: P.Cair.Mich. 3 1 (26-B9S-E) and P.Cair.Mich. 3 5 (26-B9Y-A). The script of the former is round and squat, upright but informal: its phi is disproportionately large; delta is flattened, with a wide base; serifs adorn the extremities of tau, kappa, and upsilon; and epsilon has a high middle bar. I seem to discern a high point after the verb in the third line. The poor state of preservation and the small sample size inhibits comparison, but it is not unlike P.Oxy. 80 5234 (II). The other papyrus is P.Cair.Mich. 3 5, whose upright rounded decorated capitals are akin to the script of P.Cair.Mich. 3 2, but more decorated and less regularly sized and spaced. With the exception of phi, the script of P.Cair.Mich. 3 5 is bilinear, though the letters’ modulus is not consistently square: some letters are narrow (e.g., hypsilon, alpha), while others are wide (e.g., nu, pi, kappa); some are tall (e.g., phi and beta), while others are comparatively squat (e.g., delta and sigma), as though the notional lines establishing bilinearity were inconsistently spaced. Similarly, sometimes the letters are packed tightly along the horizontal (i.1–5) and in other places more generously separated (i.6–10).
The final nine items (Nos. 17–25) were excavated elsewhere at Kom Aushim by the Michigan team. Little can be said regarding the first two. P.Mich. inv. 2755 (TM 60343) and P.Mich. inv. 2764 (TM 60688) share the field number 24-X-west side, the imprecision of which designation implies that they were ‘surface finds’ made during the initial 1924/5 season. The principals of the first season were James L. Starkey, Arthur E. R. Boak, Orlando W. (“Pip”) Qualley, and Samuel Yeivin: see Boak and Peterson 1931: 1. For the difficulties in record-keeping, see, for example, Orlando W. Qualley to Clara Petra Wollan, (15 Feb. 1925): “Yeivin had to start his plan all over again because some of the measurements were wrong.” Cf. the more expansive remarks of Orlando W. Qualley to Clara Petra Wollan (11 Feb. 1925): “… we had about the worst sandstorm of the year today. Starkey and I gave up recording on the Kom at lunch time and did some filing and cleaning at the house this afternoon. We didn’t have room to take in any more stuff until some was cleared away so it all fitted in well. P.Mich. inv. 2754 (TM 59082) bears the field number ‘24-X-west side’ in the dig’s record of excavated objects, but cf. the account of its discovery in Orlando W. Qualley to Clara Petra Wollan, (3 Dec. 1924): “Yesterday we had our first find of any consequence… On Area A, see Boak and Peterson 1931: 3–4, and Plan I.
For inv. 2755 (TM 60343), written in informal round capitals some 4 mm in height, Perrone posits a date from the end of the first to the early second century, and we concur.
Perrone 2022: 42; cf. Shelmerdine 1975: 19.
Priest 1975: 165.
Priest 1975: 165.
The field number 24-X-west side is shared by a third papyrus – P.Mich. inv. 2755a (TM 60687), which preserves a portion of For Grenfell and Kelsey’s initial inspection of Nahman’s papyri on 21 February 1920, see Todd Hickey’s forthcoming introduction to P.Mich.Cent. This lot included P.Mich. inv. 3101–3192, a group notable for the many Zenon papyri among them, of which Kondilios was a notorious source. Cf. Perrone 2022: 42. The July 1925 Kondilios lot included at least one other papyrus from Karanis: the fragments of inv. 3159 belonging to the monumental tax roll published as P.Mich. 4 224. But a multi-year interval between the acquisition of a single roll's fragments is not unusual. See Priest 1975: 124. See Figure 34.
Parallels suggest a date in the late second or early third centuries.
Shelmerdine 1975: 20, n. 4 compares Roberts 1955: no. 19b (=P.Oxy. 17 2098) and P.Oxy. 4 655 in assigning a date in the late II early III for inv. 2755a.
The challenging records from Michigan’s first excavation season complicate the archaeological analysis of another papyrus: P.Mich. inv. 2931 (24-5006-E The location of this structure in Area A is confirmed by the third plan drafted by Yeivin, titled ‘Karanis (Kom Aushim): Area A, Second Layer’, which is reproduced in Sears 2012: 174. For Area A, see the citations in fn. 67 (supra). The handwritten notebooks in which excavated objects were recorded in the first season include the names of the Egyptian excavators who dug each context, information that was lost when the notebooks were typed and, eventually, entered into a spreadsheet. For these notebooks, see Box 7, Karanis Excavation Records, Kelsey Museum of Archaeology, University of Michigan. The name is written in abbreviated form (Ah Ostman) on the page for 5006/E2 (see Figure 35), but on the third page for 5006/A, it is written in full. The date derives from Orlando W. Qualley to Clara Petra Wollan (18 March 1925), which mentions the discovery of a batch of papyri, one of which involved the selection of ephebes – evidently P.Mich. inv. 2985 (SB 4 7427), which also has the field number 24-5006E
TM Arch 90. The majority of the papyri in the batch that have been analyzed thus far date to the second or early third centuries, but there are occasional items from the first (e.g., P.Mich. 6 421; P.Mich. 9 539; P.Mich. 9 554; P.Mich. inv. 2935) and fourth centuries (i.e., P.Mich. inv. 2986, where one reads Λικιννίου Σεβαστοῦ on the final line of the side written with the fibers). 2-A). Structure 5006 is a courtyard in the so-called ‘middle level’ of Area A (on the northwest side of the 2 is a vaulted storage chamber within it. There, in March 1925, Ahmud Ostman uncovered a large batch of mixed material.2-A. See also Qualley’s letters of 10 and 11 March. It is tempting to identify the Ahmud Ostman in question with the Ahmed Osman who was a longstanding member of Petrie’s workforce: see Quirke 2010: 242.
The archaeology of Structure 5006 is obscure, and previous attempts to identify it with a property that figures prominently in the Horion archive have proven unsuccessful. See Sears 2012: 2–13, building upon unpublished research by Evelyn Adkins. The house in question, which was passed down from the family of Valeria Diodora to that of Gaius Iulius Niger, is the subject of P.Mich. 6 428 (154). So also Sears 2012: 12 n. 34. There are two examples each of Hayes ARS type 58 (24-5017B-a; 24-5016A-b) and type 61 (24-5016B-b; 24-5017B-b), and one each of Hayes ARS type 67 (24-5016A-a) and type 62 (24-5017E-b). The dates of these types are as follows: 290–375 (58); 325–450 (61); 350–425 (62); 360–470 (67). I am grateful to Nicholas Hudson for sharing preliminary data stemming from his reassessment of the Late Roman ceramics from Karanis, which will expand considerably upon the work of Pollard 1998.
The testimony of the adjacent structures’ ceramics is consistent with only the latest papyri from Room 5006E See fn. 79. On whether this papyrus belonged to a family member of Horion’s, see van Minnen 1998: 132.2. For the most part, those identified to date have been placed in the second and early third centuries,
P.Mich. inv. 2810 (TM 60520) is dated to the second century but appears a rather more refined version of the type of upright, decorated capitals preserved on P.Vindob. G 19768, See Sampson, Sawatzky, and Shields 2025: §8.
The papyrus was excavated by Hasan Osman and Aly Omar, It is tempting these excavators with the synonymous individuals from Petrie’s Egyptian workforce: see Quirke 2010: 238, 243 (for the former) and 245 (for the latter). For the former, see the record of objects. For the latter, see the caption to 5.1571.
The archaeological context of a fragment dating to the first or second century in the top layer of the site is, in a word, problematic. The situation is further complicated by the excavation’s record of objects, which notes that the papyrus was excavated “4'6" from floor in N.E. corner from filling.” Not only was the papyrus uncovered from a much later level of the site than archaeology would predict, in other words, but the circumstances of its excavation compound the difficulty. The explanation for the level and curious deposition height is doubtless to be found in the nearby activities of the sebakhin, which effectively turned the archaeological record upside down: as they dug the Other papyri were found in an adjacent room, including P.Mich. 6 418 (157 CE) and P.Mich. 6 420 (206 CE).
P.Mich. inv. 5694d (TM 66091) is a codex fragment with the field number 29-B197A-A: House 197 is located in the second level of the site, grid H11, not far from the previously discussed top-level Granary 169. There is no structure above it in the top layer.
The papyrus was dated by its editors to the third century, probably based on assumptions about the archaeology (and, in particular, the date of the B level), as well as the fact that a late second or third-century customs house receipt originates in the same context (SB 6 9234; 29-B197A-G). But the archaeology is far from consistent: the same room yielded six coins, one of which dated from the fourth year of Ptolemy II (267 BCE). The coin in question is KMA 0000.04.0099, whose field number is 29-B197A-D.
Given the structure’s location near the edge of the area entirely destroyed by sebakhin, which included the southern wall of C142K in the level immediately below, See Figure 40.
O.Mich. 3 973; (I BCE; 30-142*-E); O.Mich. 2 728 (I BCE; 30-142*-SI); KMA 0000.04.0368 (63 BCE; 30-142*-N); O.Mich. 3 1028 (I CE), P.Mich. inv. 5944 (102/103 CE); P.Mich. inv. 5939–5943, 5945 (II); O.Mich. 2 817 (II/III CE); KMA 0000.06.4988 (324/330 CE; 30-142*-N). See, for example, Sampson 2022: §6–9; Stephan and Verhoogt 2005: 196–199.
The codex fragment’s palaeography suggests its date is somewhat later than originally thought: like its editors, we characterize its script as reminiscent of the ‘formal mixed’ style, but the pronounced stylization and slope to the right indicate that we are well on our way toward the so-called ‘sloping pointing majuscule’ that arose in the later fourth century. On development of sloping pointed majuscule from the formal mixed style, see Orsini 2016: 89–93 and now Orsini 2019: 133–139. = Cavallo and Maehler 1987: no. 2b.
The next three items come from contexts in relative proximity to one another. See Figure 38.
The papyrus was assigned the field number 27-230A-H in the record of excavated objects. As is the case for other finds from House 230, this field number does not associate the house with a particular level, but the field director’s unpublished report makes clear that the excavation assigned it to the second layer of the site. It was built in part upon the foundation of C62 in layer below, See Figure 41. “None of the walls of the original C layer house, C62, alongside C56 on the west, Map 26, survived into the B layer of occupation above it, House 230. In fact, no original wall of any of the interrelated houses in that square, C62, C57, and C59, were in use in the later period, except as foundations upon which some of the walls in the B layer houses, 228, 229, 230, and 239 were built \Plate 102 and/ Map 44. Evidently there had been breaks in occupation here, times when houses in whole or in part had been unoccupied.” Peterson 1973: 446. See Figure 42. I.e., these finds have field numbers beginning 25-230 instead of 27-230: they are repeated in the entries that begin 27-230, but include there the annotation “entered in 1925–1926.” There are three examples of Hayes ARS type 58 (27-230E-lI; 27-230E-ml; 27-230E-nl), two of ARS type 67 (25-230E-f) and ARS type 91 (25-230A-l; 25-240B-d), respectively, and one each of ARS type 59 (27-230E-kl) and ARS type 93 (25-230A-m). The dates assigned to these types are as follows: 290–375 (58); 320–420 (59); 360–470 (67); 380–600 (91); 470–540 (93). It is noteworthy that one example of ARS type 91 and 93, respectively, was found in Room 230A, where the papyrus was excavated. P.Mich. inv. 4992 (27-230E-HI) was also excavated in the later, 1927/8 season: it is unpublished, but is assigned a date in the late-III/IV century by APIS, which corresponds well with the ceramic data detailed in fn. 100 as well as two coins from the reign of Constantine found in the same room (27-230E-J).
P.Mich. inv. 5760d (TM 60853) was assigned the field number 29-House NE of C65 in the record of objects. Little is known about the archaeology of this structure – which was not even assigned its own identifying number! – or its immediate surroundings. Figure 43 provides a north-facing view over the east end of C65 with large accumulations of sand and mudbrick atop it. That is consistent with the field director’s report, that it was partially covered by ancient rubbish mounds, that no remains of walls were found beneath them, and that the excavations therefore did not advance any further to the north. On the area north of C65, see Peterson 1973: 184–184.
The papyrus’ script is a small, rapid specimen of the informal round type that is notable for incorporating aspects of the formal mixed style, in particular the variation between narrow (esp. omicron, rho, and sigma) and broad letters (e.g., gamma, chi, omega). Epsilon is typically drawn in two strokes, with the second concluding in a broad extension to the right. Bilinearity is observed but for a tall iota and the hasta of rho, a letter which is also notable for its minuscule bubble. The editors offer BKT 6 II.1 (III/IV) as a parallel, but the hand of this papyrus is markedly more upright than that, and its informality makes parallels somewhat difficult to identify: it lies somewhere between documentary hands such as that of P.Oxy. 18.2183 (181) or P.Oxy. 22.2340 (192) and more formal specimens of the formal mixed style such as P.Oxy. 49 3436 (II/III). We assign a date in the late second or third century.
The final papyrus is P.Mich. inv. 6239 (TM 60527), a unique case for several reasons. For one thing, it joins with P.Aberd. 145, which was purchased on the antiquities market by James Andrew Sandilands Grant Bey, ca. 1888 but was only identified as Homer in 1991. For its identification as Homer, see Gronewald 1991: 9; for the join, see Andorlini 2001: 40–41. On the acquisition of the Aberdeen papyri, see de Ricci 1902:437; Reinach 1905: 295 n. 2; Ryholt 2019: 254. Of that group, P.Aberd. 42h, P.Aberd. 49, P.Aberd. 50, and P.Aberd. 187 have also been located in Karanis, while the case for P.Aberd. 35 is weak: see Sampson 2023: 459 n. 87. So also van Minnen 1998: 135: “The Homeric text was partly derived from below the ground level of structure 219, possibly the structure above C63.” Other papyri from below House 219 are SB 24 15895 (79 CE), an Antonine-era papyrus (P.Mich. inv. 6256), and others which date variously to the first or second centuries (P.Mich. inv. 6255), second or third centuries (P.Mich. inv. 6259, SB 22 15220), third century (SB 24 15894, P.Mich. inv. 6258, 6261, 6263b), third or fourth century (P.Mich. inv. SB 22 15791, 6263c), or fourth century (P.Mich. inv. 6262, 6263a). One is of unknown date (P.Mich. inv. 6266). The papyri from House 235 are P.Mich. inv. 6274–6290 (32-235*-RII), whose dates range from the second the fourth centuries. See Peterson 1973: 771: “None of these houses… had made use of any surviving construction of an earlier level as a base upon which to lay their walls.
The likelier explanation is that the various fragments of inv. 6239 came to light separately, in the two different findspots, and that they were assigned the same inventory number because they were promptly identified as belonging to one and the same bookroll. It is unusual for separately excavated fragments to be inventoried together immediately – I know of no other examples in the excavation data – but both finds emerged during the same excavation season in relative proximity to one another, and both the text in question and the hand (see further, below) are recognizable enough. Given that the acquisition of the Aberdeen fragment occurred decades before the Michigan excavation, the ‘wandering’ of the excavated fragments is easily explained by the previous disturbance of the archaeological record in this area of Karanis. So also Andorlini 2001: 40, n. 5.
Like that of No. 6, the hand is a fine example of a decorated rounded majuscule, but this script is even more refined than that specimen. The letters are drawn carefully, and the pen lifted consistently to make letters in separate strokes: alpha and epsilon, in particular, are in three distinct movements. Finials adorn eta, kappa, alpha, nu, and delta, and there is ample evidence of shading of strokes, with shadows along the vertical and finer hairstrokes along the horizontal axis. The middle element of omega leans backward; omicron is left slightly open at the top, as is the left side of rho’s bubble. Andorlini notes a number of parallels in assigning a date in the first or, more probably, the second century, and I concur with her judgment. See Andorlini 2001: 40.
This article has undertaken a reassessment of published Homeric papyri from Karanis on parallel (but independent) tracks, analyzing them from the perspectives of papyrology and archaeology, respectively. By considering the papyri on their own terms, it has attempted to identify new palaeographical parallels for the assignment of a date, however tentative (if not problematic) such comparisons necessarily are as regards literary papyri. On the basis of the University of Michigan’s archival maps, photographs (and captions), and associated finds in the record of objects – including ceramics – it has also attempted to outline a modest picture of their ancient contexts, as well as to reconstruct the circumstances of their deposition and persistence in the archaeological record.
A survey of the papyri that have been examined in the course of this article clusters them in various ways. In some cases, virtually nothing can be gleaned from the interrogation of archival material or archaeological legacy data beyond the basic information recorded in an object’s field number: the year of excavation, archaeological level, structure (and subdivision) of excavation, and a serialized itemization. For the surface finds of the first season (Nos. 17, 18, and 19), for example, virtually no archaeological data can be brought to bear on a papyrological analysis, but belonging to the same cluster are texts for which the legacy data is insufficient to pinpoint even the particular structural subdivision of excavation (Nos. 2–6, 9–10, 12, 15, 24). The texts of this most problematic group, notably, include papyri acquired via purchase on the antiquities market: field numbers provide somewhat more precise archaeological data than the reports accompanying purchased papyri, but neither by any means permits a rigorous contextualization. In general, there is very little overlap (or possibility thereof) between papyrological and archaeological analyses.
In precious few cases, examination of archival material sheds additional light on a text’s context. The collation of photographs, maps, and other descriptions can illustrate features of Karanis’ archaeology: not only are the findspots of Nos. 1, 7, 8, 11, 16, 20 identified, but they can also be pinpointed, visualized, or otherwise analyzed. It is worth pointing out that Vault B17F, Room B9Y, and Vault 5006-E2 all appear to be either storage facilities or immediately adjacent to storage facilities: the accumulation of large numbers of diverse papyri in two of those locations – and also in Courtyard B12L (outside, in the open air) – is suggestive of dumping.
A third cluster includes papyri from contexts that the excavation’s record of objects indicates yielded cognate finds. For these, comparison or collation of the cognate finds’ testimony with that of an individual papyrus is theoretically possible, but – a century later – only a few of the cognate object types are particularly illuminating. Papyri, ostraca, numismatics, and selected ceramics are dateable, but where comparison is possible, the results are often anomalous. Although it happened more than once at Karanis that papyri and ostraca came to light in batches, for example, examination of the assemblages to which Nos. 7, 8, 13, 14, and 20 belong indicates that such batches are heterogeneous in terms of date or content. One cannot assume that the texts have a single, common origin; it appears, rather, that the material is diverse, potentially from multiple sources, and that assemblages were accumulated or dumped over decades, if not centuries, of time. Numismatics are no more reliable: they accompany Homeric papyri only rarely at Karanis and add little to the conversation – the extreme cases of the coins accompanying No. 22 and in the vicinity of No. 23 are revealing on a different front.
The ceramics excavated at Karanis would provide the most promising basis for comparison, but the legacy data reflects the limitations of ceramology in the early twentieth century and defies a sophisticated typological analysis. Only in a few special cases of Late Roman African Red Slip fine wares can dateable types be confidently identified within the record of objects. Nonetheless, even a small sample is sufficient to demonstrate how complicated the archaeology of a given context can be; excavation discovered a papyrus from the late-second or early third century (No. 23) in the same room as African Red Slip pottery from the fourth or fifth centuries, which makes for a transparent chronological anomaly. A structure yielding fine ware pottery of similarly later vintage was located not far from another second or early third-century papyrus (No. 20). It has long been clear that the Michigan team’s approach to stratigraphy was imperfect, See, e.g., Landvatter 2014 and Landvatter 2016. So, most recently, Motta et al. 2024, contra (most notably) Boak and Peterson 1931: 5.
Archaeological anomalies can involve excavated objects or more general problems stemming from the legacy data, which brings me to the final cluster of texts. The twofold field-number assigned to No. 25, for example, suggests some sort of disturbance of the archaeological record, which scattered fragments of a single ancient book across different contexts. That disturbance is consistent with the acquisition, decades earlier, of an additional fragment of the same bookroll. Other contexts yielded finds whose archaeology is incompatible with the papyrological testimony. When a fourth-century codex fragment (No. 22) is uncovered in the same context as Ptolemaic-era coinage, or when the excavation records themselves indicate that a late-first or early-second century papyrus was found several feet above the floor level of a top-layer structure (No. 21), one must tread carefully. The former case is important for illustrating the need to conduct papyrological and archaeological analyses separately: it seems that a third-century date was assigned to the papyrus initially not on palaeographical grounds, but because of mistaken assumptions about the B layer in which it was excavated as well as the presence of a late second or third-century receipt in the same context. But the palaeography points rather decisively to the fourth century, and the archaeology is indicative of much confusion both in the analysis of layers – between B and C layer – as well as in the broad chronological range of the finds. The nearby disturbance of the archaeological record, evident on excavation maps, is likely to blame. In the latter case, the available data suggests a similar explanation: because the excavation maps indicate that the structure in question was adjacent to the crater dug by the sebakhin, it seems that their activities were effectively turning the archaeological record upside-down in this location, depositing an early papyrus from a deeper archaeological level atop material from higher, later periods. One should not be tempted to assign a later date to the papyrus by invoking its anomalous archaeology. The challenge, rather, is reconstructing how it may have ended up in the context of its excavation.
The Homeric papyri from Karanis reveal how valuable a comprehensive, interdisciplinary approach to archival material and legacy data can be – not only for our understanding of the history of both the site and its ancient readers, but also for limiting the kinds of inferences scholarship might otherwise be tempted to draw. Papyrology has a tendency to begin with the text and to proceed from there in various directions, while archaeology is wont to construct its picture from the ground up – beginning with the artifact and the context of its discovery. The two approaches can complement and reinforce one another, but equally important are anomalies or divergences in the results. Archaeology can problematize how we know what we think we know about papyri – or, indeed, whether we know it at all – and some sensitivity to the evidence’s fragility limits the scholarly impulse to extrapolate and to synthesize information in bold ways. Just because papyri are precious in the eyes of modern academics does not mean that they were handled with similar care and deliberation in antiquity, or that their ancient contexts are easily reconstructed.