Reading Scientists date Homer to 762 BCE

First they came for our linguistics, then they came for our philology.  The latest crimes against humanities have come in the form of a press release from the University of Reading where “Homer’s great masterpieces, The Iliad and The Odyssey, have been dated to around 762 BCE by new research based on the statistical modelling of language evolution.”

Wait a second…

Scientists from the University of Reading used evolutionary-linguistic statistical methods to compare the language in Homer’s Iliad with Modern Greek and Hittite (an extinct language in Anatolian branch of Indo European languages, 1200-1600 BCE) and have confirmed what many historians and classicists have long believed; that these literary classics date from the 8th century BCE.

I’m still not following…

Professor Mark Pagel’s research team analysed the differences in a common set of vocabulary items between Homeric Greek, Modern Greek and ancient Hittite and assessed the probable times in years separating these languages, given the percentage of words they shared combined with the knowledge of the rates at which different words change.  The research dated the Homerian epics with a 95% certainty within a date range of 376 BCE and 1157 BCE, with a mean estimate of 762 BCE.

Professor Pagel said: “Our analysis of The Iliad has not been informed by historical, archaeological or cultural information but by a statistical analysis of shared vocabulary between three languages and the rates of lexical replacement in Indo European languages. Yet, our estimated dates fall in the middle of classicists’ and historians’ preferred date for Homer. The outcome of this research on The Iliad demonstrates the way in which language can be used, like genes, to unravel questions in history, archaeology and anthropology.”*

Right, I understand now.  So, you’ve cooked up some lexicostatistical glottochronology  based on Bayesian Phylogenetics in order to date a linguistically non-homogenous corpus broadly within a time period from the beginning of the Greek Dark Ages down past the end of the High Classical Period, before which the epics can be demonstrably shown to have existed for centuries already.

I really don’t know what to say.  Bayesian phylogenetics applied to raw lexical data gives tenuous results at its good, bizarre BBC headlines like ‘English Language originated in Turkey‘ at its bad, and now, can be used as a terrible, terrible replacement for traditional textual criticism and philology at a bar-lowering new ugly for the mindless glottogonic speculation that is increasingly being made in this area by researchers in genetics with little to no actual historical linguistics training.

We should, however, not end this on a completely disparaging note.  One University Lecturer with whom I discussed this article today noted that it was very reassuring that the scientists had a 95% certainty that the day of Homer lie somewhere betwen 1157 BCE and 376 BCE.  More worrying is the 5% margin of error.

The article itself can be found on the Wiley Online Library, in the journal Bioessays, for those interested in reading further.

*bold emphasis mine.

Tags: ,

11 Responses to “Reading Scientists date Homer to 762 BCE”

  1. Gavin Wraith Says:

    Ever since it was “publish or perish” vendors of intellectual snake oil have been corrupting academia. The trick is to find two disparate topics and start a journal catering to the space between them: The Journal of Agricultural Psychology, say, or Proceedings of Categorical Archaeology. Biomathematics was the first of this breed. When a reviewer says “the mathematics in this article is crap” the author says “ah but you are not trained in biology”.In any case maths strikes such fear into the heart of so many people, even well educated ones,, that these shysters can get away with anything while their science-citation indices soar and their lazy employers beam with pride. I am sorry to see that palaeolinguistics has succumbed to this onslaught against scholarship.

  2. kato Says:

    Hm. Yet, their approach wouldn’t even be feasible if the principles of historical linguistics didn’t exist in the first place; talk about going in circles…

  3. Lameen Souag Says:

    The fact that this procedure came out with a result that doesn’t tell us anything new is a feature, not a bug. The press release is inexcusably misleading, but even so Pagel’s quote makes it pretty clear: his point is that, if this method gives a correct answer in this case where we already know the date, we should be able to use it in cases where we don’t know the date: say, Ainu yukars, or Bosnian epics. Of course, before attempting to do any such thing I’d want to see the method tested on a much wider range of texts; a single trial is not convincing, and, frankly, if he can’t improve on those error bars it’ll be useless anywhere – there aren’t many texts which we can’t already place to the nearest millennium! But to criticise him for coming up with a result that just tells us less than we already knew misses the point.

    • mattitiahu Says:

      Lameen, your comments are just and fair. This post was something of a rant, after all. I would like to see if the methodology *could* be profitably applied to something else, but I am still somewhat skeptical, considering the margins of error.

      But, that said, maybe the margin of error they obtained was partially because of the heterogeneity of the corpus, and perhaps better results might be obtained if they tried to go for something linguistically more uniform like, Thucydidean prose instead.

  4. eoforholt Says:

    It’s worse than the press release makes out. In the article itself, they say that their method on its own actually yields a 95% confidence interval 61-1351 BC, a 1290 year range. It’s only when they weight their model earlier (to take into account Herodotos’ mention of Homer) that they get the range discussed in the press release (which, I might note, still spans dates later than Herodotos . . .).

    Now obviously this line of research tends to go hand-in-hand with the Anatolian ‘homeland’ of PIE and the early dates associated with that. But if we take a more mainstream date for PIE, roughly 5000 BP, then with a 1290 year range the researchers would have had (slightly over) a 1-in-4 chance of encompassing the traditional date of Homer by guessing at random. I wonder if they had to run their simulation 4 times or more before they got ‘publishable’ results . . .

  5. John Cowan Says:

    I had a fair amount of parallel vocabulary, about 1500 words, for a little family of seven languages once, and I fed it through some glottochronology software just to see what would happen. Sure enough, the tree it produced corresponded to what we already believed to be the correct relationships, though the dates were based on a simplistic replacement model which we just ignored.

    Unfortunately, doubt was cast on this pleasing result rather quickly, when a colleague noted that the software had calculated the cognacy between the two most closely related languages as 104%. That was “the thirteenth stroke of a crazy clock, which not only is itself discredited but casts a shade of doubt over all previous assertions.” (A.P. Herbert)

    We never found out what went wrong (most probably a round-off error), but just dropped the results — and the software — on the floor.

    • John Cowan Says:

      I should have mentioned that the colleague in question had in fact written the software many years before, but he held no brief for the methodology.

    • David M. Says:

      It is a bit unfair to to criticise a method based on some buggy software your colleague built. The 104% glitch is a glitch in the program and not necessarily the method.

  6. kato Says:

    i’m still unsure (with regards to what lameen says) of how these kinds of processes work. in all these cases we know the initial parameters (i.e. which inputs (languages) are more archaic and which aren’t), and so of course they will tell us, at a maximum, what we already know. but what i don’t understand is how this kind of thing can be universally applicable, given that languages change at different rates? but their process *is* informed already by what they intend their outcome to be.

    • mattitiahu Says:

      Related to that point, another possible problem in this instance, is that while the Homeric language is very archaic and heterogenous, is the relative uniformity of the Greek literary language down to 1976 when Katharevousa was finally abolished, which I would expect (but am unable to prove), probably had some retarding influences on the lexical renewal of the language.

      The circumstances and attestation of the Greek language are just simply unusual compared to most other world languages.

  7. Llamabahama Says:

    Pagel and his cohorts did say that words hardly change over 15000 years (not to mention, grammatical patterns, syntax, well that’s not important anyways); needless to say, anything that Pagel claims, I will be extremely skeptical of.

    For instance, if you’re not familiar with Castilian, can you tell me what cenizas are in English? NO!? HOW DARE YOU NOT UNDERSTAND ONE OF THE DEEP ROOTED WORDS! FOR SHAME. Joking aside, it means ashes, and that one word should be enough to anyone to see how mistaken “Professor” Pagel is, that’s right I’m calling his credentials into question, big whoop, got a problem, want to fight about it?

    “You, hear me! Give this fire to that old man. Pull the black worm off the bark and give it to the mother. And no spitting in the ashes!”

    Also they seem to have forgotten how thou and ye were replaced in standard English by you, and how we have words like y’all and yinz and you lot and you guys to indicate the plural.

    Spelling-wise, me and me and me are the same, how ever an approximation between English, French, and Castilian/Italian/Various other Romance languages (including Class. Lat.) would be: mee, muh, and may. So they seem to not only discount the fact that in speech they’d be hardly recognisable, but that the words really aren’t as stable as they claim, not to mention spoonerisms becoming law in a language, bird and third were bryd and thryd (or thrid, can’t rightly remember) in Old English lord was hlaford. Sheriff was scir gerefe. (or shire reve).

    tl;dr Basically, Pagel should be charged with one count of academic fraud, one count of linguistic fraud, and one count of public indecency/public misinformation. Whuhchall thank? (What do y’all think?) Another example alone based on my normal pronunciation of English.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


Get every new post delivered to your Inbox.