I’ve just finished reading Andy Bevan’s and James Conolly’s book on their archaeological survey of Antikythera, Mediterranean Islands, Fragile Communities and Persistent Landscapes: Antikythera in Long-Term Perspective (Cambridge University Press 2013). It is a very impressive book about a very impressive archaeological project. In two seasons, they surveyed virtually the entire island (18.58 sq km, 90% of the island’s total area) with walkers spaced fairly tightly (15 m). They collected their tract data every 10 m, so every sherd collected in fieldwalking can be located within a 20 sq m strip. There is some error involved in this, but nevertheless this is, as far as I know, the best spatial control that has been achieved by any field survey in the Mediterranean.
I’m going to skip discussion of the analyses (which equally impress) because I was particularly interested in their collection strategies (for regular tractwalking, as opposed to more intensive “site” collection strategies). Fieldwalkers counted all artifacts they saw in a corridor extending 1-2 meters on either side of their line and collected feature sherds: i.e., “rims, bases and handles, decorated body sherds” (14). Now obviously such a strategy is a compromise, as all collection strategies are. Bevan and Conolly write that “we see the collection of feature sherds as a reasonably sensible fieldwork compromise” (53) and I don’t disagree. Where I do disagree is when they go on to say (in the next sentence, same page) that
Many of the proposed alternatives, such as only collecting a ‘representative’ example of each type of sherd found in the current survey unit (Moore 2008), are substantially less effective because they tend to force very high-level interpretative decisions (e.g., about what exact kinds of pottery styles and fabrics are present) to be made very early on in the field, prior to artefact cleaning and without much overview of the entire assemblage (which also leads to significant practical delays if only one or two specialists are asked to make this decision or much greater bias if a wider group of semi-trained surveyors are asked to do so).
Moore 2008 is an article by Scott Moore, a ceramicist with whom I’ve worked in Cyprus (PKAP); the article describes the Chronotype system developed by Timothy Gregory, with whom I’ve worked in the eastern Corinthia (EKAS), so I’m pretty familiar with the system. I’ve worked with it for six or seven field seasons at those projects. And I don’t think that the critique of it by Bevan and Conolly is fair.
The criticism, to recap, is that this system (and others like it) require high-level interpretive decisions in the field, prior to cleaning and without much overview of the entire assemblage. What is this high-level interpretive decision? Field walkers are instructed to pick up the first artifact they see and every other artifact that is different (with respect to fabric, surface treatment, body part, and thickness), and that if they are uncertain whether an artifact is different, to collect it.
Now this doesn’t seem to me to require much high-level interpretation. My experience is that it’s pretty straightforward. The walkers don’t need to know “what exact kinds of pottery styles and fabrics are present” nor do they need to have an “overview of the entire assemblage” (Bevan and Conolly 2013, 53). They just pick up what they see, and if it might be different, they collect it. The experience of everyone who has used the Chronotype system (and this is also based on experimental data) is that field walkers overcollect, i.e., collect redundant data, so that if the field walkers make any mistakes, it is in the direction of total collection. (This is a summary of Gregory 2004: 28-29, published here and cited elsewhere by Bevan and Conolly).
The other issue is that the identification of what is and what is not a feature sherd is a kind of decision (high-level or otherwise), so that the “feature sherd” collection strategy is open to the same criticism: i.e., that it forces field walkers to make a decision before the artifact has been cleaned. I suppose that a critic would point out that it’s easy to identify handles and rims and bases, and I suppose that it is. But it’s not clear to me — maybe it’s clear to everyone else — that it’s so much easier than what Chronotype asks you to do. The other issue is what happens when a fieldwalker is uncertain? In the Chronotype system, the impulse drilled into fieldwalkers is “when in doubt, collect.” What if, using the “feature sherd” strategy, a sherd looks like a plain body sherd, but in actuality preserves a tiny portion of the rim that goes unnoticed? Perhaps the fieldwalker would collect it anyway. But I would guess that in most cases, s/he would not.
In any case, this brings us to the weakness of “feature sherd” collection strategies: they do not collect sherds that are potentially diagnostic. As Bevan and Conolly repeatedly point out, fabrics are important indicators of chronology, function, and even origin, and they are becoming more and more important over time, not less. I point this out because as Bevan and Conolly (52) correctly observe, the
significant discrepancy between what we now know and what we might know in <the> future, is one of the reasons we feel strongly that surface survey, like other kinds of archaeological fieldwork, should make systematic, permanent artefact collections …*
All this is not to say that the Chronotype system is perfect: it is certainly not. In fact, if I were doing survey in Antikythera, I would not use it. So I am certainly not questioning their decision to collect feature sherds. After all, the proof of the pudding is in the eating, and there is no doubt that Bevan and Conolly have provided us with rich and nourishing results (to extend the metaphor rather awkwardly). But neither do I think that the critique of Bevan and Conolly is fair, nor do I think that we can solve all of our problems by collecting feature sherds only.
The problem here is rather more complex, I think, and it has to do with research design, a theme that featured prominently in an article that I co-authored with Bill Caraher and David Pettegrew in 2006. Selective collection (i.e., collecting feature sherds, and especially the Chronotype collection method) is a way to reduce the collection of redundant data. On the Eastern Korinthia Archaeological Survey (EKAS), for instance, the project counted 146,599 artifacts in only 3.85 sq km. The Antikythera survey, in contrast, counted 66,000 artifacts in 18.58 sq km. That is a huge difference! It is somewhat larger than it seems, of course, because EKAS used a tighter 10 m walker spacing as opposed to Antikythera’s 15 m, but if we correct for that, we will see that EKAS counted over 7 times as much pottery as Antikythera did. What about collection? EKAS collected 38,337 artifacts (26%), Antikythera collected 25,675 (39%) by my count of their open data, which I downloaded here.
It’s probably impossible to find a place to do survey in Greece that is more different than the Corinthia than Antikythera (or the other way around, depending on your perspective!), and so the differences in the collection strategies aren’t that surprising. The Corinthia is a high-density artifactual landscape, even for Greece. Antikythera is very low-density, in part because (a) there are virtually no rooftiles (my dream!!!) and (b) for significant periods of its history, the island was abandoned or had a very low population. So a survey in Corinthia needs to find ways to cope with the fact that a high-intensity collection strategy would have produced insane numbers of artifacts. This is simply not a concern of a survey in a place like Antikythera (or at least, it’s a pretty minor concern). All of us, I think, would agree that it makes no sense to sample a high-density artifactual landscape the same way you would sample a low-density artifactual landscape. The lower the density, the more sense it makes to move towards total collection and towards piece-plotting individual finds.
In sum, I’d argue that the Chronotype collection strategy, flawed as it may be, is one way to deal with a particular problem. I’d readily concede that it may not be the best. On the other hand, it does do a good job at (a) reducing the collection of redundant data, (b) collecting useful qualitative and (to a lesser extent) quantitative data about the surface assemblage, and (c) avoiding the potential biasing effect of ignoring non-decorated body sherds that can, despite their humble appearances, be highly diagnostic (my ceramicist friends insist that they can do a lot with body sherds, and I believe them).
This is not to say that Bevan and Conolly did the wrong thing: I don’t think that at all. But I do think that we need to weigh the advantages and disadvantages of our methods with clarity. This is something that is a real strength of the Antikythera survey book, in fact. Their second chapter ends with a page of meditations on “methodological limitations.” So it is a minor criticism to pick them up on their brief discussion of collection strategies. My point is not to nitpick, but to extend the discussion about survey collection a bit further, beyond the text of their immensely rewarding book.
* In the … I have omitted reference to the same article by Timothy Gregory (2004), which advocates for processing artifacts in the field, although it should be pointed out, as Gregory indeed does (2004:30), that this procedure was instituted by necessity, due to restrictions placed on the survey in the archaeological permit.