You never really think about it until you reach a certain level of decision-making. You go out, conduct your survey, and then there are sites. Contracting issues aside, you’ll eventually return to those sites for evaluation, or maybe you’ve already written them off.
But there’s one nagging issue that I’ve discussed with quite a few younger archaeologists. Site boundaries don’t reflect anything. Oh sure, we use GPS data of the surveyed finds and relate that data against the landscape, but how do you know where the boundary should fall? How do you know when a bunch of positive shovel tests should be two sites and not one?
The short answer is: we don’t. Seriously.
Sites aren’t homogenous things. They’re full of empty spaces where we don’t find any evidence of human activity. Maybe the activity avoided those areas. Maybe plenty happened there, but no evidence was detectable.
People tend to clean their places.
I’d like to think that you could look at archaeological and ethnoarchaeological evidence to develop a model to predict the distance where clusters of artifacts would be independent from each other. You can’t, though. To do that, you would have to know information that could probably wouldn’t know at the survey level of archaeological work. How many people lived there? What were they doing there? How long did they intend to stay there? How long did they actually stay there? All of these are factors on the size of an occupation.
THEN, you’d have to consider that the site isn’t a single occupation, but is instead a palimpsest of occupations. These occupations are probably separated by hundreds of years, at least. They’re independent of each other, which means there’s no possible way to predict how far apart they are from each other. That is, if they don’t overlap at all.
See, the problem is that the site is a flawed concept that doesn’t work well as an analytical unit. The variability between sites can be too great to make them comparable as units. Sites make great managerial units, however, and allow us to designate areas to be studied or protected. It’s important to recognize that the boundaries are just a handy fiction, however. We’re just drawing a circle somewhere in the landscape where the data has run out. That’s fine, because we need (or at least really really want) to designate boundaries so we can manage them and one place with no data is probably just as good as another place with no data. The important thing is to do your best ensure that the actual archaeological remains are encompassed by the boundary and that your reasoning for placing the boundary is explicit.
Most of the CRM reports that I’ve read (and perhaps a few that I’ve written) contain absolutely no discussion on why the investigators decided that the site boundaries were put where they were put. Sometimes it’s obvious because it’s a road or the edge of a survey, but other times it’s not. If it’s not made explicit, there’s no possibility of a reasoned approach towards the treatment of the site’s fringes. Everyone will have to assume that the site goes all the way to the edge, whether or not it does.
There are numerous scholarly articles about this in the literature. I recommend reading Dunnell’s The Notion Site, if you can find a copy.