I’m not planning to blog a lot on the Astronomical Orientation of Ancient Greek Temples as is openly accessible. Your comments are going to carry a lot more weight there than here. But I’ll try and keep track of what other people are saying elsewhere. I’m expecting this to be the first paper of a developing argument, so I’ll need to see what people identify as problems and address them. There’s two comments in the Times today which I think neatly highlight one of the issues. One is from Efrosyni Boutsikas and the other from Mary Beard.
Boutsikas’ objection is interesting. I wanted a comparison data set to examine for Greece and the only one I know of that’s published is Retallack’s in Antiquity. To be honest it’s not entirely fair to use Retallack’s data as he wasn’t that concerned with astronomy. Instead he was looking at geomorphology, and I think he has some really interesting results. Now Boutsikas has her own survey, which she did specifically with archaeoastronomy in mind. That’s why I’m interested when she says of 107 temples in Greece only 58% faced east. That might cause me some problems and here’s why.
First we need three variables. n is the number of temples. That’s 107. p is the probability and event will happen, and q the probability it won’t. p+q = 1 because something has to either happen or not happen. So what value is p? It depends on what Boutsikas means by east. If she means the eastern half of the horizon, then p is 0.5 and so is q. By pure chance we’d expect np temples to face east. That’s 53.5. In her survey 62 temples faced east. That’s more than chance, so I’m right yes? Not so fast.
There’s never going to be exactly 53.5 temples facing east. Around 19 times out of twenty 20 there’ll be 53.5 plus or minus two standard deviations. If you want the chance of a false positive to be lower than 1 in a 100, then you need 53.5 plus or minus three standard deviations. The standard deviation (σ) for this kind of distribution is calculated by the formula:
σ = √npq
If we want the one in twenty confidence that means
σ = √(107 × 0.5 × 0.5) = 5.2
So 19 times out of 20 you might expect to find between 43.1 and 63.9 temples facing east. The result of 62 is inside this range, so there’s nothing significant. How do I explain that? I’m not sure I can. I can’t say what time period her temples come from. If she’s looked at Bronze Age sites and Roman sites in Greece then we’re not comparing like with like and it’s possible that when we compare temples built in the archaic and classical periods as they were in Sicily then we’ll have more of a match. Another possibility is that I’m simply wrong.
But this depends on Boutsikas meaning eastern half of the horizon when she says east. I use different definitions of east for different tests and make clear which are which in the paper because it makes a difference. If Boutsikas has put her temples into four categories, north, east, south and west, then east means eastern quarter of the sky and the equations turn out differently.
np = 107 × 0.25 = 26.75 temples
σ = √(107 × 0.25 × 0.75) = 4.48
If east is the eastern quarter then 19 times out of twenty at most we would expect at most 35.71 temples. By chance there’s be over 40.19 temples in the sample less than 1 time out of 100. 62 temples would be over seven standard deviations away from the expected result. If that’s by chance it’s an amazing freak result. It means I can’t respond to Boutsikas’s claims until I can see the data to analyse, so I know what east means. She might have proven my paper wrong, or else proven it very right. How can that be when only just over half of the temples face east?
Imagine you’re at a casino. Someone is spinning the roulette wheel. Half the time it lands on the numbers 0–9 and the other half it lands on another number. It doesn’t matter than you can’t predict exactly where the next ball will land. In the long term that casino will lose money because something is affecting the wheel. This isn’t about having a hard and fast law for astronomical alignments. It’s about whether or not a significant number of temples are aligned to the sun. If you’re going say that there is or isn’t a significant number, first you have to know what a significant number would look like. Typically in the social sciences that would np +/- 2σ. I prefer np +/- 3σ because I’m making claims which people might not be comfortable with, so it’s reasonable I should provide stronger evidence.
If I am right that doesn’t mean Boutsikas’ and Retallack’s surveys can be junked. In fact it means the opposite. In Retallack’s case he’s showing there’s a clear correlation between the dedication of a temple and the soil type it’s built in. Now if there’s a general rule that Greek temples face east, the temples which don’t become more interesting because then you can ask “What’s special about those temples? Why were they built that way?” It’s the same for Boutsikas’ data. If there’s nothing special about the alignments then temples which don’t face east are nothing special. If, using this method, her data shows a tendency for eastern alignments then she has a data set with plenty of interesting temples that could tell us something about Greek religion. For instance it could highlight where a local cult was doing something special that you wouldn’t find elsewhere in Greece.
Clearly Boutsikas’ objection is serious and I’ll need to consider it carefully, but in this case it could be a case of cross-wired. I don’t think she’d seen my article when she talked to the Times because I hadn’t emailed it to her till last night. We’ve both been working on similar topics and so could have come to the same conclusions. If we been talking with each other then there could have been a bit of friction if we saw our ideas in each other’s theses. She’s been put on the spot reacting to a paper which she probably hasn’t read, but she’s clearly an expert in the subject because of her own research. She’ll go where her research takes her and I’ll go with mine. The reporter has picked up on that confusion. Does that leave him in the dog house? Definitely not.
I’m really pleased with the way Mark Henderson has written this up. It’s not his job to preach my wonderfulness, it’s to report on how this research fits in with other research. Getting the quotes from Efrosyni Boutsikas was brilliant because it shows there’s currently two models which come to opposite conclusions. As we both publish more those models will get fleshed out and adapt. Which one will be accepted? Hers? Mine? Some kind of hybrid, or even neither? It’s not just about getting the right answer. At the moment we might not even agree on what the right answer will look like. Which brings me to Mary Beard’s piece.
I think it’s great commentary. I think she’s spot on when she rejects the idea of a modern astronomy in the ancient world. I would quibble with her rejecting astronomy for the rhythms of day and night. That sounds astronomical to me and there’s also evidence the seasons were important. I think she might be trying to emphasise the importance of cosmological features, in the sense of natural order, rather than strict observation. The only real puzzle is that she’s saying that there’s it’s obvious that Greek temples align east-west when in the column next to her Efrosyni Boutsikas is saying they obviously don’t. This is a bit of an interdisciplinary gap.
From the outside you might expect archaeologists and classicists to talk to each other. They’re dealing with the same people in the same time period. In reality this doesn’t always happen. A few years back the Roman Archaeology Conference, the big conference for Roman archaeologists held once every couple of years was scheduled opposite the Classical Association conference. The two sides don’t always talk to each other. In the past few years Boutsikas has been publishing on her work. Ioannis Liritzis and Helena Vassiliou have been arguing that Greek temples were aligned towards or away from aurorae or stars. This is having no impact amongst classicists. Equally I can’t just turn around and say “Greek temples tended to face east” because all the researchers who disagree could ask “How do you know?” It’s obvious doesn’t work as an academic response, even though I agree with Mary Beard. I don’t imagine that would be her response in an academic forum. But what she’s done is she’s very helpfully shown that if I want to talk to classicists then show why I think I have something to talk about.
That’s why I’ve had to write this paper. I want to write more, but the first question anyone can ask is “How do you know that’s not just a chance result?” That’s why I developed this method. I wanted something simple and effective. The reason I put it in PLoS One is that it also has to be accessible. I’m planning to write more articles for specialist journals, but people reading those will need access to my data and my methodology. That needs to be available to classicists, archaeologists, astronomers and anyone else with an interest.
You can read the original research for free and download it at PLoS One. If you leave comments there then they’ll be seen by everyone else who examines the paper. If you’d like to blog about the paper there’s a collection of photos from Sicily at Flickr with a Creative Commons licence.
Salt, A. (2009). The Astronomical Orientation of Ancient Greek Temples PLoS ONE, 4 (11) DOI: 10.1371/journal.pone.0007903