March 3rd, 2013
In “Linguistic Evidence Supports Date for Homeric Epics” Eric Altschuler and colleagues bring “formal statistical modeling of languages” to bear on the question of when the Iliad “was produced.”1 Such an approach might prompt some interesting discussion, but only when informed by relevant expertise. Regrettably, they willfully disregard alternate forms of expertise:2
Our analysis is not informed or constrained in any way by historical, cultural or archeological information about Homer or his works…
It is disconcerting to find scholars so dismissive of pertinent information. Had they consulted with historians, classicists, or archeologists, they might have avoided some of the questionable assumptions that undergird this essay.
The authors assume that the text we have today reflects a particular, definite, discoverable moment in the past. But what moment, if any, does it reflect? The Iliad was composed and transmitted orally long before it was first written down. We have little to no evidence for how the written text varies from the oral tradition that preceded it. The oldest written fragments of the Iliad that survive date from centuries after we think the poem was first committed to writing, and the oldest surviving complete copy is the fabulous Venetus A, which dates from nearly 1800 years later. Therefore, on the basis of the written text alone it is difficult to know what moment if any it reflects.
The authors also assume that the text and modern Greek have varied independently as a function of time and only time. They exclude all complicating factors that might affect both the text itself and the development of modern Greek. That assumption might work for culturally neutral texts, but it seems problematic for more important texts. The Iliad was a culturally dominate text that served as the basis for education for generations. Authors consciously imitated Homeric expressions precisely because they were thought to have been used by Homer.
Altschuler and colleagues are following in the footsteps of earlier researchers who have oversimplified a problem so that they can apply models from the sciences to questions of cultural production and transmission. Those simplifications weaken the argument on which their conclusions rests.3
In the end they offer two conclusions:
our analysis of common vocabulary items in the Iliad increases our confidence in its age and shows how even fictional texts can preserve traces of history.
Neither conclusion is particularly revelatory.
Their analysis suggests as most likely a date of early- to mid-eighth century, depending on how they tweak their model by including or excluding information. Their model does not, however, rule out as impossible dates that seem unreasonable, dates as early as 1351 BCE and as late as 61 BCE (when they chose to include obviously relevant historical evidence, the range narrows to 1157-376 BCE). Archeological and textual evidence suggests that such dates are impossible—archeological evidence suggests that Troy was destroyed in the early-12th century BCE (see here for more on Troy), and Plato quotes Homer in the 4th century BCE.
Their other contribution is equally banal. Historians and textual scholars have known for centuries that “even fictional texts can preserve traces of history.” That fact is one of the cornerstones of textual criticism. Assuming that all texts preserve history, five and a half centuries ago Lorenzo Valla demonstrated that the Donation of Constantine was a late forgery. Historical criticism continues to accept as a truism that all texts preserve history.
Unfortunately, the authors seem to have contempt for other scholarly but non-science disciplines that could have contributed to and improved this project.4 They denigrate the conclusions derived from historical and archeological sources as “historians’ and classicists’ beliefs.” Labeling something a belief characterizes it as opinion, as view, as conviction. Beliefs are not rational and grounded in evidence. Likewise, for historians and classicists there is a “preferred date for Homer.” Like belief, prefer reduces the claim to opinion. In contrast, and despite the absurdity of some of the results offered by their model, they give a “formal quantitative estimate” and offer a “prediction” and estimate “with 95% confidence intervals.” Their model “returns a date for Homer.” Unlike the amateurish methods and conlcusions of historians and classicists, Altschuler and colleagues offer results that overcome the stain of opinion. They are scientific conclusions grounded in a “Bayesian approach” and backed up by 95% confidence intervals.
It is too bad they didn’t consult with classicists, historians, and perhaps archeologists. If refined and developed in consultation with relevant experts, their model might be able to offer interesting insight into historical questions. As it stands, however, they seem to have squandered considerable money on something that doesn’t contribute anything of value.
Thanks to Brett Mulligan for offering his expertise and keeping me from making egregious errors.
1 By “produced” it seems they mean something like first written down, though they slide unhelpfully between various ambiguous expressions such as “the age of the Homeric epics,” or “mean estimate for the date of Homer’s works,” or simply “a date for Homer.”.↩
2 The authors are not alone in denying relevant expertise. In November Gerald Crabtree, a geneticist at Standford, admitted his lack of expertise in a field and then went on to speculate about “one of the most important questions” in that field.↩
3 See John L. Cisne’s “How Science Survived: Medieval Manuscripts’ “Demography” and Classic Texts’ Extinction,” Science 307 (2005): 1305–07 and the various responses to it. (Unfortunately, all are behind Science’s paywall).↩