Scientists prove Beowulf and the Iliad are true stories! Not. Sometimes scientists can be so clueless, you just want to pat them on the head and go “Aw, that’s so sad.” To get up to speed on this new silliness, check out John Bohannon’s article for Science Now: Is Mythology Like Facebook?, which summarizes this scientific paper: Ralph Kenna and Pádraig Mac Carron, “Universal Properties of Mythological Networks,” Europhysics Letters 99.2 (July 2012) #28002. To be fair, they only claim to have evidence “the societies” and “some of the events” in them are true, not the entire stories as wrote. But really they don’t.
Kenna and Mac Carron mapped out the social networks in three myths (Beowulf, the Iliad, and the Táin, a lesser-known Irish epic) and tested those networks for the properties of real networks. Then they used as “controls” four works of modern fiction (Les Misérables, Shakespeare’s Richard III, The Fellowship of the Ring, and the first book in the Harry Potter series). (If you are noticing a fatal flaw here already, you get ten bonus points. Fifty points if you already notice more than one fatal flaw.)
Facebook is a real network that’s been well-studied this way (hence the title of Bohannon’s article). And real networks have certain properties. As Bohannon explains:
For example, one universal feature of real social networks is that they are highly clustered, with tight clumps of people who all know each other. These groups are linked to each other by highly social people known as “connectors.” Real social networks also have a property called “small world,” which indicates that there is never more than a few degrees of separation between any two people. Such statistical properties have been found in networks of movie actors, jazz musicians, and even scientific collaborators. If the ancient myths were based on real people, Kenna and Mac Carron expected to find the same patterns.
They did. The character social networks of all the stories, including the modern fictions, were “small world” and highly clustered. But the modern fictions differed from the ancient myths, as well as from real social networks, in several ways. For example, in the fictional narratives, most of the minor characters link to the main character. But that wasn’t true for the myths. If the researchers removed the character of Beowulf from the network, they found that the other characters are linked with one another in other ways. “Certainly there are similarities between fiction and real life,” says Kenna, but in fiction “everyone tends to be connected to everyone else, otherwise the story becomes too hard to follow.”
That last sentence sounds like a contradiction, but I think a connecting sentence must have gotten lost. That last sentence refers to a different property of fictional networks, which is “small world without clustering” or what the authors called in the paper “too small world,” in which every character is connected to many other characters, independently of the main character, which is not typical of real networks. Thus we have two measures of artificiality: one in which everyone is connected only or mostly through one or a few key persons, and one in which everyone is connected to everyone else almost regardless.
What’s wrong with this? Lots.
#1: Scientists Need to Learn How Historical Fiction Works
The first problem here should be obvious. Bohannon claims earlier on, of Beowulf and the Iliad, that “historians and archaeologists agree that much of those ancient narratives was based on real people and events.” Even on the most charitable reading, this confuses the difference between history and historical fiction.
Certainly Beowulf is based on real historical persons and contexts (Wikipedia has a nice summary), but Beowulf himself is by all accounts a fictional character inserted among them, to play out a supernatural drama that obviously has no basis in history (no monster was terrorizing the lands that required finding a special young warrior named Beowulf to gather a band of men to hunt it down). This is actually admitted by Kenna and Mac Carron, so I should not take them to task for that (as I did in the original draft of this critique). But the general fact of it illustrates that any character could be similarly inserted, possibly thereby undetected. This tale appears to be a Nordic (and possibly Christian-influenced) adaptation of Virgil’s Aeneid. Like the Aeneid, if in Beowulf any ancillary (non-central) character was invented to fill out the story, would Kenna and Mac Carron’s method be able to detect it? I don’t see how. We can only confirm the historicity of its characters by other, more traditional methods; and by that method, we have.
But as much cannot be said of the Iliad, in which none of the characters have any likely basis in history. There was probably indeed a Trojan War, between the Ilians and an alliance of Achaeans (proto-Greek nations that sailed and lay siege in Ilia, probably in defense of their coastal holdings in Turkey). But reality almost certainly ends there. Achilles, Hector, Odysseus, Ajax, these are again inventions of religious fiction, inserted into a distant memory of historical events. The war was certainly not fought over Helen. And even if there ever was a king Agamemnon, he lived centuries earlier (and thus was in no way involved in the Trojan War).
Indeed, the Iliad was written over the course of several centuries. For example, the story randomly and inexplicably jumps from bronze-age technology to iron-age technology and back again throughout, indicating that the tale as a whole was freely built up over a very long span of time. Yes, that means a man named Homer did not write the Iliad, at least the Iliad we have, but rather a series of anonymous bards over many centuries did. So it would be absurd to suggest the characters in it are likely to be real, much less its stories about them.
This may even explain the findings of their study: the heroes of the Iliad all derive from local national myths from diverse Greek nations, which may have been added into the story over time, to create an increasingly panhellenic epic as bards traveled the Greek states. Each local myth involved its own artificial family and political social network built around the central hero or heroes (usually mythical royal houses). Then by assembling all these heroes onto one battlefield, they brought their independent networks with them. The result would look very much like Facebook. By accident.
These kinds of possibilities (and I am sure I will have not exhausted them all in this essay) are precisely why the mathematical properties of the networks are meaningless by themselves: you have to ask a Classicist (especially an expert in Homeric studies) to look at the nodes of the network that make it look real and tell you the most likely literary explanation for those connections being in the story. “Not written by a single author,” “artificial panhellenic epic built over centuries,” and so on (more examples below) might turn out to be the better explanation.
#2: Scientists Need to Learn Logic
The methodology in this paper is also illogical. It employs one giant fallacy of affirming the consequent. To illustrate, I’ll adapt a table I used when identifying this same error in the writings of Bart Ehrman:
|If p, then q.||If it’s a real social network, it has properties [x, y, z].||If a dog ate your homework, then you have no homework to turn in.|
|q.||The social network in story P has properties [x, y, z].||You have no homework to turn in.|
|Therefore, p.||Therefore the social network in story P is a real social network.||Therefore, a dog ate your homework.|
Really, you can prove anything with logic like this. Which is precisely why you can prove nothing with logic like this.
The study does attempt to avoid this fallacy by using a scientific “control,” but the control they chose is wholly invalid. In effect, they want to argue that “If it’s a fake social network, it will never have properties [x, y, z],” and they try to do that by testing an arbitrary and unscientific selection of modern works of fiction, and used them to represent all fake social networks, even those constructed in ancient literature. This is just error piled on top of error, one for every word I put in bold.
First of all, the study’s authors fail to consider that fake networks can have the same properties as real ones, especially in cultures that create myths to map and mimic social realities, which was often the very purpose of ancient myths. Those social realities included extended families and community-pervading political relations, two things often lacking in modern novels because most people don’t live in those social networks anymore. Hardly anyone is connected to a political figure, and extended families are rare. Yet those kinds of social networks were so common in antiquity they were emulated in their myths, deliberately. Indeed, in a way, myths were training for the tracking of social networks (thus, as discussed in Rafaella Cribiore’s Gymnastics of the Mind, students in ancient schools were often tasked with mapping the social networks in them, for example being forced to memorize the families and genealogies of famous heroes like Ajax, along with lists of their colleagues).
Obviously, such mythical networks will have “properties [x, y, z]” even when fabricated. Because they were designed to be. And even when they weren’t designed to be, they still reflected the social realities familiar to authors and readers, which realities were very “Facebook-like” (in the complexity of political and familial relations binding nations and communities together). Thus ancient myth and modern fiction are simply not sufficiently similar for this study to use one as the control for the other.
Secondly, just arbitrarily picking some works of modern fiction is the height of unscientific claptrappery. Can you imagine doing a scientific control study by arbitrarily picking four people to be the control group? Not only is “four” way too few to be representative, not only is the selection not a random one from among all available works of fiction, and not only are the genres chosen not functionally similar (the author of a play has specific interests in mind as far as plotting characters that can be cast and represented conveniently on a stage, concerns which are not shared by a novelist; likewise, Harry Potter is about the students, faculty, and alumni of a single elite school, whose social network properties are not going to look like Facebook even if it were real), but I’ll bet any literary expert could point them to confuting counter-examples.
For example, what do you think will happen when they test The Godfather for social network properties? I suspect it will look a lot like Beowulf. Which means, a lot like Facebook. Because the author of The Godfather was aiming to mimic exactly that kind of network. The same might be true of War and Peace. Likewise, The Belgariad, a complete work of fiction (not even historical fiction), also might have those features, because its authors liked to play with complex social networks.The same might be true of Game of Thrones. Even if these don’t fit the analysis mathematically (I am only guessing they could), there still may be plenty of works of fiction that do. You can’t just claim there aren’t, because you picked four books out of a bin.
And again, that’s still modern fiction. That is no parallel to ancient myth, either in cultural context or aims and function. So even if they constructed a proper control (which they didn’t), it still would not be a valid control.
#3: The Result Is Useless in Application
For these reasons we can’t conclude a mythical hero existed because the social network he is inserted into looks real. Not only because real networks are easily mimicked (as is almost certainly the case in the Iliad)–especially by persons aiming to mimick such networks, and persons in cultures that live and breathe such networks and depend on navigating them–but also because fictional persons can easily be inserted into real networks (as is almost certainly the case in Beowulf). Thus their tool is useless.
This kind of social network analysis can’t be used the other way around, either. Because a historian might deliberately focus on a single individual and his connections (e.g. Lucian’s account of Alexander of Abonuteichus), and thus leave out the social network attributes, even though the person in question really existed (as we have effectively confirmed archaeologically). Or he might focus on people who are all connected to each other (and thus not connected through only one or a few central characters). Either way, looking exactly like the works of fiction Kenna and Mac Carron analyzed (which means even their premise “if it’s a real social network, it has properties [x, y, z]” fails to relevantly hold, because the premise “all accounts of real social networks will have properties [x, y, z]” is false, and they are analyzing literary accounts of networks, not the networks themselves). And importantly, not all historical works will operate the same way (some will capture real networks, some will focus on single-characters and who they connect with, some will focus on highly interconnected cabals–and therefore no one set of “network properties” will characterize “histories” as histories).
For both reasons, we cannot use this tool in the myth-or-historicity debate surrounding Jesus, as some have suggested in various places online. Jesus would almost certainly fail the analysis (removing him from the network would leave few connections among the characters in the Gospels, making the network look highly artificial), unless we allowed network connections to be drawn among the historical actors (e.g. Herod, Pilate, Caiaphas) using data outside the Gospels, but that would just confirm the obvious: that some historical persons were used as backdrop for the mythical tales of Jesus. No one doubts that. What we want to know is whether Jesus (the story’s Beowulf or Odysseus) existed. And this tool cannot help us with that.
Thus, if this analysis were turned on the Gospels and showed a seemingly-artificial Jesus-centric social network, that would not mean the Gospels were fiction. Nor would the Gospels showing a more distributed network mean the Gospels were fact.
[Update: those interested in pursuing the Beowulf case will find even more reason to doubt its historicity from the analysis and bibliography provided in Eve Siebert, “Monsters and Dragons and Dinosaurs, Oh My,” Skeptical Inquirer 37.1 (January/February 2013): 43-48, albeit there dealing with an entirely different claim.]