Did an ancient Greek play really feature Kannada dialogue?
Reviewing scholarship of the Charition mime found in the Oxyrhynchus Papyri
Recently, I came across a rather intriguing reel on Instagram by an account called @historyofbangalore:
The reel is about Papyrus Oxyrhynchus 413, one of the papyrus manuscripts from a collection that was discovered by Bernard Grenfell and Arthur Hunt at an archeological site in Oxyrynchus, Egypt in the 1890s. This Greek manuscript, dated to the 2nd century CE (i.e. when Egypt was part of the Roman Empire), contains a stage play that appears to be a low-brow comedic adaptation of Iphigenia in Tauris by Euripedes. Interestingly, the Oxyrynchus adaptation is set in India, and is about a Greek girl who is held captive by an Indian king, and then freed by her brother and his search party who outsmart the king by getting him drunk.
According to the reel, the 2nd century play features dialogue in the Kannada language, which was first recognized by Eugen Hultzsch, the German Indologist. This is an incredible claim, as it predates the oldest known Kannada inscriptions (the Halmidi inscription) by about three centuries. Specifically, the reel claims that Hultzsch recognized the following line as Kannada:
bēre konca madhu pātrakke hāki - “put the booze in another cup”
Now, this sounds a bit too good to be true, especially since we know that in the first millenium, Old Kannada (haḷe-gannaḍa) was quite different from the the modern version of the language. It seems implausible that a Greek play from the 2nd century features a form of Kannada that would be readily recognizable to speakers of the language today.
To get to the bottom of this, let’s first try to find the original manuscript, which is stored in the Bodleian Library at Oxford. Fortunately for us, a high-resolution image of the manuscript (both sides) is freely viewable on the Bodleian Library’s website. I’ve cropped out full quotation containing the line mentioned above:
This is a bit hard to read, so we can refer to the transcription as it appears in The Oxyrynchus Papyri: Part III, written the discoverers themselves (Grenfell and Hunt) and published in 1903.
Despite being written in Greek script, this words here do not make any sense in Greek. As noted by Grenfell and Hunt (and reproduced in their transcription), there is actually a little dot between each word in the original manuscript (p. 43). Interestingly, these dots are only present in this particular passage. The fact that the word breaks are clearly delimited constrains the translation effort quite substantially, which will be important in the discussion that follows. If, like me, you’re not super comfortable in reading the Greek alphabet, here is the text transliterated into Roman letters as part of the English translation, a few pages later in the same book (this passage was obviously untranslatable to Grenfell and Hunt):
Note: For simplicity, I will use the Roman script for transliterating both Greek and Kannada words in the remainder of this post. Also, I’ve never studied ancient Greek, so I will have to rely on the translations provided by the authors of the papers we will examine (who I assume were all well-trained in the Classics).
From the transliteration by Grenfell and Hunt, we see that the actual quote referred to in the Instagram reel is:
bere konzei damun petrekio paktei
Which is a bit different from what the reel claimed, which was:
bēre konca madhu patrakke hāki
What the reel reported on was Hultzsch’s interpretation of the original line in the manuscript. Based on Grenfell and Hunt’s 1903 edition of the manuscripts, he published his own article the following year titled “Remarks on a Papyrus from Oxyrhynchus.” He provides a bit of justification, but not much, for each word:
bere = bēre (“different”), which he notes is “still in daily use” in Kannada.1
konzei = koñca (“a little”), which he says is an “equally common” word
petrekio = pātrakke (“to the cup”), the dative form of pātra (“cup”)
damun = madhu (“wine”), which he simply claims is a transposition of d and m
paktei = hāki (“having poured2”), apparently “perhaps an incorrect rendering”
Hultzsch thus translates this part as “having poured a little wine into the cup separately.”
There are actually several more non-Greek words and sentences throughout the play, several of which Hultzsch tries to interpret as Kannada. To strengthen his claims, he points out that some of the Greek characters in the play apparently translate what the Indians are saying to their compatriots, which semantically constrains the phrases. Here are the other phrases Hultzsch picks out, along with his interpretation of their meanings:
Line 59: brathis = bēr āḍisu - “let us3 play separately.” In the play, the apparent Greek translation provided by the character “B” for this utterance is “let us draw lots for the shares.”
Line 66: kottos zotit4 = kuḍisu jhaṭiti - lines up with the apparent Greek translation provided by character “C” as “give to drink quickly.”
Line 83: panoumbretikatemanouambretououeni = pānam bēr etti kaṭṭi madhuvam bēr ettuvenu (“having taken the up the cup separately and having covered [it], I shall take wine separately”). Here, Hultzsch hypothesizes that manouam is a clerical error for madouam, from which he gets madhuvam.
There are a few more non-Greek phrases in the play, for which Hultzsch admits that he is unable to generate plausible Kannada translations.
To me, the most problematic aspect about Hultzsch’s interpretation is that he does not discuss Old Kannada anywhere at all; it’s not clear that he even knows about attested forms of Kannada from premodern times. He seems to be exclusively trying to fit each word into an acceptable modern Kannada form. In particular, he considers paktei to be an “incorrect rendering” of hāki, but does not discuss the fact that Kannada underwent a p→h sound change between ancient and modern times, which may actually strengthen his claim.5 Still, Hultzsch’s Kannada interpretation of the play’s “barbarian” dialogues are intriguing since their content does make some sense in the context of the play - a scene in which the king and others are drinking.
The British orientalist L. D. Barnett published a rebuttal of Hultzsch’s claims in 1926, titled “The Alleged Kanarese Speeches in P. Oxy. 413.” First, Barnett is skeptical of the unsystematic manner in which Hultzsch modifies the original spellings of the Greek words to obtain his translations, most egregiously the transformation of damun to madhu. He also points out that the oldest Old Kannada text, the Kavirājamārga,6 is from the ninth century, or seven centuries after the Oxyrynchus manuscript, and so we simply do not have a good reference for what 2nd century Kannada was like. This is the basis of his strongest criticism of Hultzsch: “if a proposed reconstruction agrees with the rules of the oldest classical texts, we may admit it, but only provisionally and with reservations, and if on the other hand it shows features of the medieval or modern dialects we must unhesitatingly reject it” (p. 13). Along these lines, Barnett makes the following objections:
koñca and hāku are relatively modern Kannada words, and koñca would require a case-ending in Old Kannada.
ettuvenu, the first-person present-future conjugation of ettu (“to rise”), would not have existed in Old Kannada - it would have instead been ettuvem.
The Sanskrit words pātra and jhaṭiti are out of place for ancient colloquial Kannada (although he points out that a derived form, jaḍiti, is attested in medieval Kannada).7
The second-person imperatives kuḍisu and bēr āḍisu are supposedly rendered by the Greeks as kottos and brathis, omitting the final -u sound. According to Barnett, Old Kannada would only permit consonant endings if they were nasals or liquids.8
Barnett also takes issue with the way Hultzsch contextualizes his Kannada interpretations within the play.
After the king says zopit, B asks “what do they say?” C responds, “give them a drink, quick.” Thus, it’s unclear if “give them a drink quick” is the in-play translation provided for the phrase kottos zopit, or just zopit.
As previously mentioned (footnote 3), bēr āḍisu is a 2nd person imperative, not 1st plural imperative. Taken naturally in Kannada, this phrase would mean “play separately,” or, according to Barnett, “put into play a change or difference.” But the Greek translation provided is “let us share in the parts.” Thus Hultzsch’s interpretation is a bit of a stretch.
Barnett also argues that Hultzsch’s interpretations are rather forced and nonsensical:
pānam bēr etti kaṭṭi madhuvam bēr ettuvenu - this is just an awkward sentence - Barnett says, “Seriously speaking, we cannot understanding the meaning of Hultzsch’s translation.”
The use of bēre is quite frequent in Hultzsch’s translation, even though it often seems out of place. Barnett notes that according to Hultzsch, the Indians in the play “use it in every phrase, and it never seems to have much sense.”
Barnett concludes by saying that the non-Greek passages of this play are equally likely to be gibberish or some Indian language, either Indo-Aryan or Dravidian, “but if it is an Indian dialect it has yet to be interpreted.”
Overall, I think that Barnett’s objections to Hultzsch are pretty solid. However, we could cut Hultzsch a bit of slack by considering that the Greeks writers may have simply been doing a bad, corrupted impression of Kannada based on overhearing their Indian colleagues. This would be like if you forced me to write some dialogues in Spanish right now (with no Google Translate!) - I would probably mess up the grammar quite badly, and overuse a small number of words that stick in my mind from my few interactions with Spanish speakers.
In 1936, the Mangalore-born Indian historian Bhasker Anand Saletore published a book called Ancient Karnataka, vol. 1: History of Tuluva, in which he also argues that the non-Greek passages of the Charition play are indeed Kannada. Interestingly, he proposes entirely different translations than Hultzsch, which I assume reflect actual knowledge of Old Kannada. However, Saletore violates the word delimiters in the “bere konzei…” passage, which one could argue makes his interpretation unconvincing.9 He also goes one word further in this passage than Hultzsch:
bere konzei damun petrekio paktei kortames = bēre koñca īyada munna bētir ēkeyo bhāga tekoḷ tammā īsu, which he translates as “Why did you put down your cup before some more was served? Take a portion! Serve a little, Brother!” (p. 594). Saletore goes on to ambitiously interpret every single non-Greek passage into what I imagine is some form of Old Kannada. It’s all quite intriguing, but at some point one has to wonder if this whole exercise has devolved into witty wordplay rather than linguistics scholarship.
In 1991, Richard Salomon (professor of Sanskrit at the University of Washington) revisited the disputed interpretations of the Charition play’s supposedly Indian-language passages as part of a larger article entitled “Epigraphic Remains of Indian Traders in Egypt.” While he contends that Barnett’s arguments against Hultzsch were strong, the idea that the passages contain some impression of a Dravidian language should not be discounted entirely. As Barnett notes, “it has long been known…that the Roman imperial sea trade with India was centered in south India” (p. 735). In the context of other epigraphic discoveries in Egypt - a Prakrit Brahmi tablet of likely south Indian origin, as well as pottery with two proper names written in Old Tamil script - it seems highly likely that a community of South Indian merchants lived among the Greeks in the area. Thus, it is plausible that the writers of the Charition play really did know a Dravidian language, or had friends who did, and that they weren’t just writing pure gibberish. In this context, Salomon calls on Dravidian specialists to revisit Hultszch’s claims, even if they are flawed.
In a 1993 addendum to that article, Salomon acknowledges further scholarship on the Charition play that he had been unaware of two years earlier - the 1936 work of Saletore, which we discussed above, as well a 1985 piece by P. Shivaprasad Rai ambitiously titled “Sariti: A 2000 year old bilingual Greek-Tulu play.” I haven’t been able to find this latter paper, which apparently claims that the language is Tulu, not Kannada, and that the name of the princess, Charition, is actually a Hellenized version of Sariti.
In the end, it seems that the verdict is still out on whether the non-Greek passages of the Charition mime are Kannada, some other Indian language, or pure gibberish meant to represent what Greeks thought that Indians sounded like. While the Instagram reel was thus a bit misleading, I have to credit it with bringing the existence of this manuscript to my attention in the first place. At any rate, the play serves as a reminder of how internationalized the ancient world was, with Greek, Roman, and Indian merchants interacting with each other on a regular basis. More personally, this exercise has reminded me that I ought to learn a bit of ancient Greek and Old Kannada at some point.
βραθις!
Hultsch actually refers to the Kannada language as Kanarese, as it was known in English at the time.
Note that Hultzsch interprets hāki as an absolutive, while the Instagram reel interprets it as a polite 2nd person imperative (these forms are identical in modern Kannada).
In modern Kannada, āḍisu is actually the 2nd person casual imperative (“you, play!”) and not the 1st person plural imperative (“let us play”) as Hultzsch seems to think. That form would be āḍisōṇa. This really calls into question how much Kannada Hultzsch really knew when he wrote the article…
Actually, this word is transcribed as zopit (originally ζοπιτ) in Grenfell and Hunt, but with the note that the letter π is indistinct in the original manuscript. Thus, Hultzsch feels justified in replacing it with τ to form ζοτιτ.
To be clear, I myself have not studied Old Kannada, so I don’t know for sure if the old form of the verb hāku was actually pāku - I’m just assuming here.
As previously mentioned in this post, the Halmidi inscription predates the Kavi-rāja-mārga by about four centuries and is the oldest known example of the Kannada language known so far. However, it was not discovered until 1936 - a decade after Barnett’s article. In any case, I imagine that the short inscription probably doesn’t really provide a comprehensive picture of the intricacies Kannada language of its time.
Personally, I am skeptical of this argument - Barnett seems to think that Sanskrit words would have been rare in 2nd century Kannada, which is not obvious to me.
In modern Kannada, dropping the final vowel of a word is common in colloquial speech, but from my observation, it can only happen within a sentence (i.e. between words), and generally not at the end of a sentence/utterance. For example, you may hear the sentence nīnu āḍisu bēku pronounced as nīn āḍis bēku (“you should play”), but you wouldn’t hear someone just say ādis! (“play!”) - rather, they would say āḍisu! or āḍiso!
Or, one could hypothesize that the dots between words were just there to help the Greek actor break down the syllables easily, even though the actual underlying language broke down the words in different places. But this scenario sounds rather contrived.