 So I want to talk about something that is a bit of an inconvenient truth in Roman studies and that is the amphora reuse problem. Amphorae are the main type of evidence that we use in Roman archaeology to talk about long distance transport foodstuffs, many olive oil, wine and fish products. And they are also one of the most used comparable sources of information to talk about centuries long change in the functioning of the Roman economy. Yet those amphorae could be reused. And that fact at best has been ignored in Roman studies and at worst has been declined the night that it even happened. So I think we should investigate this issue a little bit more. Now I'm talking about a very particular type of reuse. This defined as type A reuse by Teodor Peña. So reuse involving an application similar to the vessel's prime use application without any physical modification to it. So basically a pot that was filled with olive oil at some point and that then was reused because it no longer had olive oil in it. It had wine in it, for example, or fish product, these kind of things. So if an amphorae had olive oil in it in a first instance, which is called the prime use content, and then it is emptied of its content, cleaned and refilled with the same type of product, olive oil perhaps from a different source, doesn't matter, that is not the kind of reuse I'm talking about. It is the kind of reuse where we can identify that the vessel was emptied of its prime use content and then a different type of food stuff was put into that amphorae. And related to that is the assumption that every one of these amphorae have a sort of principal content, that an amphorae type, so a particular shape of amphorae produced in a particular region, was also linked to a particular type of food stuff from that region. So for example, amphorae produced in North Africa could be related to locally produced olive oil. And so should we interpret the distribution of these amphorae, equaling the distribution of those products? This is something that we do all the time in Roman archaeology. When we excavate these amphorae, when we analyze selected context, and we look at the distribution of amphorae containers, we assume very often that they had a principal content, that an amphorae type is related to a particular food stuff product, and that we can look at something that we could call market shares of these products, that African olive oil had this particular distribution in the Roman Empire, and that changed in this way and it's directly correlated to the African olive oil amphorae. So something like that is implied here. These are selected context from Rome and Austria for Italian wine amphorae, and it just details proportions of the amphorae and the origins of these amphorae, but it is very tempting to interpret that as, so there was so much of Teranian wine consumed, so much of Adriatic wine consumed, so much of this wine consumed. We don't always make these assumptions, but they are kind of there, and we are not addressing this problem. Now, I do think it is a problem, and I want to use the example of African one distribution. This is a very badly photographed distribution map of African one distribution, but it was the only one I could find in a library when I didn't have a scanner in my pocket. And this is just a typical distribution map that we see all the time and that we like to think we can actually interpret. It just shows red dots that are places where African one was found and excavated in the 1970s. It doesn't matter. This is an archaeological distribution map of amphorae. Does this equal the distribution of olive oil that was also produced in present day Tunisia? Now, there are selected archaeological contexts that suggest that might not be the case. So this is the content of the Grado shipwreck, a shipwreck full of these kind of African amphorae, two hundreds of them. They are typically associated with olive oil, and African one is one of those types of amphorae that are particularly well studied for which we can never know 100% for sure that it was just exclusively used for African olive oil, but we've got a pretty good hunch that we assume it's olive oil that was in there. And in this case, it's full of 10 to 11 tons of fish products. Now, these are pretty big volumes, and even though this particular shipwreck might just be a one-off, or it's definitely an anecdotal evidence, there's not a lot of great context slice shipwrecks where we could potentially identify these kind of things. It does suggest that it was possible that we should be cautious, and at the very least we should try to evaluate the impact of reuse on amphorae distributions. Because if the impact is bigger than we think it is, then it changes the way in which we need to look at distribution maps for understanding the Roman economy. Then the distribution of amphorae types really can't be equal to the distribution of the foodstuffs. There's actually, besides just studying these fantastic contexts, there's a variety of ways in which reuse could be evidenced. A lot of them could be from the type of sites, so there's places, maybe perhaps more in port sites, riverine or maritime ports, where we could have refilling stations where the ships could arrive with the amphorae being typically transported in ships. They're a bit awkward for overland transport, so perhaps we can assume a higher probability of reuse at ports. They could be emptied there, and there could be infrastructure for cleaning the amphorae and for refilling them with a local product or whatever product should be available there. Storage facilities for empty amphorae that are saved in order to refill them. But if you invest in infrastructure like that, it basically implies that you don't have enough capacity or ability to produce amphorae yourself in the local region. So what are those places where that is just not a thing, where there's no tradition of producing these amphorae locally, where maybe primary resources are not available to produce amphorae, to the scale that is required for their use as containers. We can look at the vessels themselves, see if there's use where that is indicative of reuse, or whether there's repair that is indicative of an amphorae being emptied and then being refilled, or whether an amphorae is stopped with something, with a type of stuffer, so like something to close the amphorae, that is not actually from the same location as the amphorae itself was produced. Beautiful rare evidence is written stuff, so some of these amphorae had inscriptions on them in the form of titillipicti, and in perhaps some cases that have been identified they could be overlapping. Titillipicti that could indicate the prime use content, and that an overlapping written description could indicate the reuse content. But these are extremely rare. Now, there's a lot of ways in which we could explore this. The problem of amphorae reuse is that it's very difficult to tie down exactly how it happened. Theories about where and when and in which proportions it happened vary a lot. So there's a lot of things that interact there. On the one hand there's just our theories about how the Roman macroeconomy functioned, and the implications for reuse there. So what was the role of imperial regulation of trade and taxation, and how did that structure, the main long-distance trade flows, and the need for amphorae reuse? There are certain types of amphorae for which this is absolutely nonsensical to assume this happened to any scale, such as the Dresel 20 olive oil amphorae produced in the south of Spain, and then sent in mass mainly to the capital of Rome, where we have a huge mountain called Mont de Testatio, mainly of these Dresel 20 amphorae, so they were just literally emptied and broken. Not saying that this happened for those amphorae, but there's other types out there. So the difference between the type of site, and whether it happened more often at riverine or seaports, and the distance from those ports, the way in which the Roman transport system could have structured the distribution of the amphorae, and it's driven out of foodstuffs, and the need for places that are not well connected in the Roman transport system to be dependent on reuse amphorae or not. Basically, loads of theories. It's actually a really challenging problem, and honestly, when I first heard about it, I was kind of like, yeah, let's not address this, let's just keep it an inconvenient truth. But I find this, well, two things. On the one hand, it's a really important problem to tackle, because even though it's possible and very likely that for some amphorae and in some contexts, the impact was totally minimal, there is indication that in some context it has a huge impact. So we need to be able to at the very least discard the issue of reuse. And secondly, it's totally doable. On the one hand, the theories are modellable, and there is evidence that we can collect in order to validate our formal approaches for evaluating the impact. So I think this is a challenge that we should just try to overcome. What I want to present in the rest of my presentation is a proof of concept. So set your brains to 100% fiction mode. There is not really a lot of archeological data that is going to be involved in what I'm going to say now. I just want to give an indication that the formal simulation side of what I think should happen is possible, and that if we can combine that with a very large scale campaign of residue analyses and integrating large data sets of ceramic distributions and so gathering the empirical evidence that's needed to validate those models, then we can come up with a way of evaluating the impact of Roman amphorae reuse better than we are able to do now. I'm just going to talk about simulation now. So the first aspect of this proof of concept model is to have a transport system over which I will simulate the distribution of these amphorae and the foodstuffs. The only digital open access model of the Roman transport system that's available now is the Orbis model, which is extremely coarse-grained and is really just useful for proof of concept and general trend research. But it's accessible. Then there's two key theories that I will explore. On the one hand, differences between theorized probabilities of reuse. Some people who deny it happens say there's a 0% probability of reuse. This type of amphorae was never ever ever reused. And if you find one where it's evidence, then you are lying. Or there's 100% reuse. It's really, really frequent. It happened all the time, which is something that almost no one says. And then there are people who are a bit more pragmatic. People like Teodor Penya, who published this model of live cycle of amphorae, who argues that in some cases like African one, he theorizes an 80% probability of reuse in the port of arrival. And for amphorae in general, a 38%. That's the first set of experiments. And the second set of experiments focuses at reuse at ports. So it is not crazy to assume that reuse might happen with a higher probability at port sites because they were mainly meant for maritime transport. I explored conditions in which there's 0% reuse at ports, 50% and 100%. Now, I will also have... I need to implement a basic macroeconomic model of how the Roman economy functions. So we have some production in North Africa, focusing on this case of African olive oil. And we have a distribution of demand for all of those settlements in the Orbis network. And the proxy evidence I use for demand are just population estimates which are derived from the work of Jack Hansen. And I'm just using a correlation of population with demand. That's it. Now, what I do is I then... I create a simple model. I simulate distributions of these amphorae and I simulate the distribution of the food stuff. I can keep track of both of them. And I compare the outputs of the simulations with four different types of data patterns that are suggested as indicative of this African one distribution. So... And this is the only bit where archaeology comes in. So, you know, there's no special collected data or anything. It's just squinting your eyes at this map. The focus of this distribution is on the Western Mediterranean. So we expect more of this distribution in the West than in the East. It reached as far as Britain, maybe in small volumes. We definitely got there. There are high volumes of prime use contents, so not the reuse in Rome and Austria. And there is a high proportion of reuse at Aquilea. Now, that is just... We have the Grado shipwreck, which is close to Aquilea. And so I want to check in which conditions I have a high proportion of reuse at the Grado shipwreck. How can I reproduce that? That's like just one anchor point that I want to use. The model is very simple. Initially, I create a demand for a product, which is correlated to the population. I generate a supply, which is the production. Then there's a discard of that production, which is a proportion taken from Ted Penya's model. Then amphorae are sent to places that still have an unsatisfied demand for products. Then depending on the settings of the experiment, reuse selection of amphorae will happen. So this could be at a port and then there's a 50% probability or 100% probability depending on the experiment I'm testing. This is where reuse happens. Then all the amphorae that are transported in that turn are not selected for reuse are deposited. So this is how I get distributions of pots and also distributions of foodstuffs, simulated ones. In the next time step, the ones that are selected for reuse will be further disseminated. Here are just some example results of one experiment. Experiment number six, in which we assume 50% reuse selection at ports. We have to modify from 25 to 50% the production capacity of the producer here. So basically their ability to satisfy all the demand in the system. What I want to show here is that on the left-hand side we have distribution, simulated distributions of the pots. On the right-hand side we have simulated distributions of the olive oil. In this theoretical scenario, the distribution of the pots is different than the distribution of the olive oil. The cause of that is the theorized impact of reuse. So I'm going to just read out key conclusions of this proof of concept work. Under no circumstances that I tested in which I had zero or 100% reuse, did I get anything that was even remotely close to the four archaeological markers that I tried to reproduce. So this is almost anecdotal, but the extreme scenarios we can discard very easily and provably with a method like this. But given that we all agreed might have probably happened to some extent, what can we say about the more moderate proportions of reuse? The cases of 18% to 38% reuse, combined with more frequent reuse selection at the ports, are more promising. So these simulation experiments here tend to compare more favorably with the archaeological markers that I'm using to validate these models. In no cases where there is reuse is the distribution of the container the same as the distribution of the foodstuff. And so if we even hypothesize a small effect of the reuse, we expect there to be a difference between those two distribution patterns. So we cannot assume that if we look at a distribution of pottery, we are looking at a distribution of foodstuffs. But this is very important. In all of these scenarios, I can explore in which context this impact is bigger. Is there a higher probability of reuse to be resulting in assemblage in a port site, in a riverine site? What is the difference with places that are far away from this port site? What is the typical assemblage at small terrestrial sites, for example? So the proportion varies greatly depending on the context and generally following this logic. Higher proportions of reuse are simulated for assemblages further away from the producing region and higher proportions of reuse are simulated at terrestrial sites further removed from ports. Now, what it doesn't succeed is representing the more limited presence of African one in the eastern Mediterranean and for that I think in just future proof of concept models I should have an element that represents the structuring of state demand for goods both for the capital of Rome, the large consumption centers but also the troops along the limous. Now, finally, I just want to conclude that all of this is fiction but I think this is something we should do more of. I think I can show that simulating these theories is possible and useful. There's three useful things I think this does. On the one hand we can have the simulated geographical distributions implied by our theories. Secondly, we can estimate proportions of reuse for site assemblages under a particular context. We can use those estimates to calibrate the amphorae that we excavate. This would be hugely useful to be able to have like a probability range like an error margin of everything we excavate. And then to identify differences in the reuse proportions between these site contexts. We need to combine these kind of things with large scale campaigns of residue analysis and with bringing all the amphorae data that we have available to us together and I do think that is possible right now. So anyways, just the proof of concept thing please do talk about this with me because it's obviously an early stage. Thank you.