The Big Filter: Archaeological Site Files and Site Formation Theory in the Duck and Lower Tennessee Rivers

30 Days of Tennessee Archaeology 2016, Day 17

Shane Miller
Mississippi State University

In my first year of grad school, I remember reading several articles by a guy named Michael Schiffer. Most archaeologists are familiar with him – he’s a prolific writer who’s spent a large portion of his career writing about “site formation theory,” or more broadly, what are the behaviors that cause archaeological sites to form, and what are the kinds of processes that bias how we find them.

It’s no stretch to say that now most archaeologists constantly ask what are the cultural processes (i.e. c-transforms) and natural processes (i.e. n-transforms) that are biasing, or filtering, our data?

A few years later, I decided that I’d like to analyze the distribution of archaeological sites from central Tennessee as part of my dissertation research, and I approached Suzanne Hoyal, the Site File Curator for the Tennessee Division of Archaeology, for help. And she said, “That’s a big area, and the coverage is really uneven” – in other words, some areas had seen a lot of research, and others very little. She suggested that I stick to the Lower Tennessee River and the Duck River. Because of WPA and TVA projects like the Kentucky Lake, Columbia, and Normandy Reservoirs, those areas have seen a lot more work than a bunch of the surrounding areas.

After walking out of her office, it hit me that I should probably look at the entire research area like it was a giant archaeological site.


The distribution of recorded archaeological sites in the Duck River Drainage.

So, I spent the next two weeks compiling information on sites, and then the next several months tracking down as many publications and site reports as I could get my hands on. Why? Because, the first issue Suzanne told me to expect was that many of the sites files haven’t been updated since they were first reported. When all was said and done, I had compiled a database of 2,211 sites.


The frequency of sites in the Duck River drainage by date reported from the Tennessee Division of Archaeology site file records.

Next, I noticed that in some places, like the Lower Tennessee River, many of the sites were recorded in the 1930s before archaeologists had figured out the projectile point chronology for the Mid-South. So, I spent a lot of time going through boxes in the McClung Museum “typing” the projectile points recovered from these sites.

Then I noticed a couple expressions of the “Teotihuacan effect.” Imagine you’re doing an archaeological survey of the Valley of Mexico. One of the first sites that you’re going to record would be the biggest, like Teotihuacan. In the Lower Tennessee and Duck Rivers, sites that have either earthen mounds or shell middens have lower site numbers, meaning they were recorded first. In other words, because they’re highly visible, they are more likely to be identified and recorded.

Another bias I noticed was what I call the “vacuum effect.” Through time, the frequency of lithic scatters with an unknown cultural affiliation increases, while the number of sites with projectile points that are distinctive enough that we can “type” them diminishes. This pattern is likely the result of collectors happening upon sites and removing the points while leaving the rest of the assemblage behind. It’s like the projectile points have vacuumed off the sites.


The frequency of temporal components in the Duck River drainage. Note the large number of sites with “unknown” cultural affiliations.

These were just a few of the distortions to the data with which I had to contend. I also had to look out for and deal with:

  1. temporal biases (more recent sites are more likely to be found),
  2. landcover biases (artifacts are more likely to be found where there’s open ground and plowed fields),
  3. geomorphologic biases (artifacts are more likely to be found where there’s shallow, stable surfaces), and
  4. geologic biases (sites are more likely to be found near sources of chert, or flint).

To some, this barrage of biases would appear overwhelming. However, I feel like I share the position of most archaeologists in that these challenges are what make working with archaeological data so much fun. I’ll spare everyone the gory details of how I managed to determine the extent to which these biases affected my sample of sites, and just get straight to the punchline: lurking behind all of these biases was a pretty robust pattern.

During the initial colonization of the Lower Tennessee and Duck Rivers, sites were more frequently located in the area that is now Kentucky Lake. Then, as climate began warming in the Early Holocene, they begin to be creep up to higher elevations, most likely as deciduous forests also moved up to higher elevations. Finally, on the eve of the Mid-Holocene, sites are found everywhere, and it stays that way for the remainder of the Archaic period.


Variation in elevation across the Duck River drainage (vertical exaggeration 10x).

Thankfully for me, Suzanne directed me to one of the few regions where this pattern was likely to “pop,” and just as importantly there was actually good enough data to tease out the effect of biases that filtered the data before it ever made it into my database.

Why does this matter? Eastern North America is one of at least six places in the world where plants were domesticated. In fact, the Hayes site, where the earliest date for domesticated sunflowers has been documented, is located along the Duck River. Currently, there is a debate as to whether the people who domesticated these plants did so out of necessity. In other words, were there too many people relative to the availability of food resources? This study supports the conclusion that the Duck River was effectively “packed” in the millennia prior to the appearance of domesticated plants, and that perhaps necessity was the mother of invention.